Download poxygit, a debugging git server, and clone from it using NTLM,
both IIS-style (with connection affinity) and Apache-style ("broken",
requiring constant reauthentication).

Attempt to obtain a Kerberos ticket from LIBGIT2.ORG and then clone the
Negotiate-protected site at test.libgit2.org with that ticket.

Our docker builds are getting expensive, let's cache some of this.

The valgrind in the PPA is broken and ignores `--exit-errorcode`.
Build and install our own.

Use a multi-stage docker build so that we can cache early stages and not
need to download the apt-provided dependencies during every build (when
only later stages change).

Deleting the apt cache can be helpful for reducing the size of a
container, but since we don't push it anywhere, it only hinders our
ability to debug problems while working on the container.  Keep it.

The lsb-release command is missing on our images; just show the
information from the file instead of relying on it.

Provide a sane default for `CMAKE_GENERATOR` in the build script so that
it can be invoked without having to set that in the environment.

Right now, all tests in libgit2's CI are being executed as root
user. As libgit2 will usually not run as a root user in "normal"
usecases and furthermore as there are tests that rely on the
ability to _not_ be able to create certain paths, let's instead
create an unprivileged user "libgit2" and use that across all
docker images.

While we were still supporting Trusty, using Ninja as a build tool would
have required us to first setup pip and then use it to install Ninja.
As a result, the speedups from using Ninja were drowned out by the
time required to install Ninja. But as we have deprecated Trusty now,
both Xenial and Bionic have recent versions of Ninja in their
repositories and thus we can now use Ninja.

The Valgrind version shipped with Xenial has some bugs that keep our
tests from working due to bad interactions with openssl [1]. Fix this by
using the "hola-launchpad/valgrind" PPA that provides a newer version
for which the bug has been fixed.

[1]: https://bugs.launchpad.net/ubuntu/+source/valgrind/+bug/1574437

Pass the flag "--no-install-recommends" to apt-get in order to trim down
the number of packages installed, both reducing build time and image
size. As this also causes some required packages to not be installed
anymore, add these explicitly to the set of packages installed.

Reformat both Xenial and Bionic's Dockerfiles to use best practices.
Most importantly, we now run `apt-get update` and `apt-get install` in
one step followed up by removing the package lists to speed up
installation and keep down the image size.

While Xenial provides libssh2 in its repositories, it only has version
1.5.0 available. This version will unfortunately not be able to connect
to GitHub due to their removal of weak cryptographic standards [1]. To
still enable our CI to execute tests against GitHub, we thus have to
update the provided libssh2 version to a newer one.

Manually install libssh2 1.8.2 on Xenial. There's no need to do the same
for Bionic, as it already provides libssh2 1.8.0.

[1]: https://github.blog/2018-02-01-crypto-removal-notice/

We're about to phase out support for Trusty, but neither Bionic nor
Xenial images provide the mbedTLS library that's available in Trusty.
Build them for both to pull them in line with Trusty.

The Coverity build is still referencing an old "trusty-openssl"
container that is not provided by either our own now-inlined images nor
by the libgit2/libgit2-docker repository.

Convert it to build and use Xenial images instead.

Support for the LTS release Ubuntu 14.04 Trusty has been dropped in
April 2019, but Azure is still using Trusty as its primary platform to
build and test against. Let's deprecate it in favor of Xenial.

The Docker images used for our continuous integration builds currently
live in the libgit2/libgit2-docker repository. To make any changes in
them, one has to make a PR there, get it reviewed, re-build the images
and publish them to Docker Hub. This process is slow and tedious, making
it harder than necessary to perform any updates to our Docker-based
build pipeline.

To fix this, we include all Dockerfiles used by Azure from the mentioned
repository and inline them into our own repo. Instead of having to
manually push them to the CI, it will now build the required containers
on each pull request, allowing much greater flexibility.

On Win32 builds, the PID file created by git-daemon contained in invalid
PID that we were not able to kill afterwards. Somehow, it seems like the
contained PID was wrapped in braces. Consequentially, kill(1) failed and
thus caused the build to error.

Fix this by directly grabbing the PID of the spawned git-daemon process.

On Win32 build hosts, we do not have an SSH daemon readily available and
thus cannot perform the SSH tests. Let's skip the tests to not let Azure
Pipelines fail.

Right now, we maintain semantically equivalent build scripts in
both Bash and Powershell to support both Windows and non-Windows
hosts. Azure Pipelines supports Bash on Windows, too, via Git for
Windows, and as such it's not really required to maintain the
Powershell scripts at all.

Remove them to reduce our own maintenance burden.

Until now, we always had the CC variable defined in the build.sh
pipeline. But as we're about to migrate the Windows jobs to Bash, as
well, those will not have CC defined and thus we cannot use "$CC" to
determine the compiler version.

Fix this by only executing "$CC" if the variable was set.

We currently specify the CMake generator as part of the CMAKE_OPTIONS
variable. This is fine in the current setup, but during the conversion
to drop PowerShell scripts this will prove problematic for all
generators that have spaces in their names due to quoting issues.

Convert to use an explicit CMAKE_GENERATOR variable that makes it easier
to get quoting right.

We're about to phase out our Powershell scripts as Azure
Pipelines does in fact support Bash scripts on all platforms. As
a preparatory step, let's replace our MinGW setup script with a
Bash script.

Azure Pipelines supports bash tasks on Windows hosts due to it always
having Git for Windows included. To support this, the Git for Window
directory is added to the PATH environment to make the bash shell
available for execution. Unfortunately, this breaks CMake with the MinGW
generator, as it has sanity checks to verify that no bash executable is
in the PATH. So we can either remove Git for Windows from the path, but
then we're unable to execute bash jobs. Or we can add it to the path,
but then we're unable to execute CMake with the MinGW generator.

Let's re-model how we set the PATH environment. Instead of setting up
PATH for the complete build job, we now set a variable "BUILD_PATH" for
the job. This variable is only being used when executing CMake so that
it encounters a sanitizied PATH environment without GfW's bash shell.

Since we have migrated to Azure Pipelines, we have deprecated and
subsequentally removed all infrastructure for AppVeyor and
Travis. Thus it doesn't make a lot of sense to have the split
between "ci/" and "azure-pipelines/" directories anymoer, as
"azure-pipelines/" is essentially our only CI.

Move all CI scripts into the "azure-pipelines/" directory to have
everything centrally located and to remove clutter in the
top-level directory.

Right now, we have an awful hack in our test CI setup that extracts the
test command from CTest's output and then prepends the leak checker.
This is dependent on non-machine-parseable output from CMake and also
breaks on various ocassions, like for example when we have spaces in the
current path or when the path contains backslashes. Both conditions may
easily be triggered on Win32 systems, and in fact they do break our
Azure Pipelines builds.

Remove the awful hack in favour of a new CMake build option
"USE_LEAK_CHECKER". If specifying e.g. "-DUSE_LEAK_CHECKER=valgrind",
then we will set up all tests to be run under valgrind. Like this, we
can again simply execute ctest without needing to rely on evil sourcery.

We currently have no job that compiles libgit2 with the WinHTTP backend
for SHA1. Due to this, a compile error has been introduced and not
noticed for several months. Change the x86 MSVC job to use the HTTPS
backend for SHA1. The x86 job was chosen with no particular reason.

The interactions between `USE_HTTPS` and `SHA1_BACKEND` have been
streamlined. Previously we would have accepted not quite working
configurations (like, `-DUSE_HTTPS=OFF -DSHA1_BACKEND=OpenSSL`) and, as
the OpenSSL detection only ran with `USE_HTTPS`, the link would fail.

The detection was moved to a new `USE_SHA1`, modeled after `USE_HTTPS`,
which takes the values "CollisionDetection/Backend/Generic", to better
match how the "hashing backend" is selected, the default (ON) being
"CollisionDetection".

Note that, as `SHA1_BACKEND` is still used internally, you might need to
check what customization you're using it for.

Explicitly enable the `builtin` regex backend and the PCRE backend for
some Linux builds.

This reverts commit 1fe3fa5e.

Enable hard deprecation in our builds to ensure that we do not call
deprecated functions internally.

Like 811c1c0f, disable the SSH tests on
macOS until we can resolve the newly introduced infrastructure issues.

Coverity is back but it's only read-only!  Agh.  Just allow it to fail
and not impact the overall job run.