The function to write trees allocates a new buffer for each tree.
This causes problems with performance when performing a lot
of actions involving writing trees, e.g. when doing many merges.
Fix the issue by instead handing in a shared buffer, which is then
re-used across the calls without having to re-allocate between
calls.

odb: only provide the empty tree

stransport: make internal functions static

diff: document `git_diff_from_buffer`

Only provide the empty tree internally, which matches git's behavior.
If we provide the empty blob then any users trying to write it with
libgit2 would omit it from actually landing in the odb, which appear
to git proper as a broken repository (missing that object).

SecureTransport: handle NULL trust on success

The `SSLCopyPeerTrust` call can succeed but fail to return a trust
object if it can't load the certificate chain and thus cannot check the
validity of a certificate. This can lead to us calling `CFRelease` on a
`NULL` trust object, causing a crash.

Handle this by returning ECERTIFICATE.

clar: fix parsing of test suite prefixes

Patch image initialization

odb: freshen existing objects when writing

Since writing multiple objects may all already exist in a single
packfile, avoid freshening that packfile repeatedly in a tight loop.
Instead, only freshen pack files every 2 seconds.

When writing an object, we calculate its OID and see if it exists in the
object database.  If it does, we need to freshen the file that contains
it.

sysdir: don't assume an empty dir is uninitialized

Fix outdated comment

Enable https transport for custom TLS streams

Don't try to determine when sysdirs are uninitialized.  Instead, simply
initialize them all at `git_libgit2_init` time and never try to
reinitialize, except when consumers explicitly call `git_sysdir_set`.

Looking at the buffer length is especially problematic, since there may
no appropriate path for that value.  (For example, the Windows-specific
programdata directory has no value on non-Windows machines.)

Previously we would continually trying to re-lookup these values,
which could get racy if two different threads are each calling
`git_sysdir_get` and trying to lookup / clear the value simultaneously.

mwindow: init mwindow files in git_libgit2_init

When passing in a specific suite which should be executed by clar
via `-stest::suite`, we try to parse this string and then include
all tests contained in this suite. This also includes all tests
in sub-suites, e.g. 'test::suite::foo'.

In the case where multiple suites start with the same _string_,
for example 'test::foo' and 'test::foobar', we fail to
distinguish this correctly. When passing in `-stest::foobar`,
we wrongly determine that 'test::foo' is a prefix and try to
execute all of its matching functions. But as no function
will now match 'test::foobar', we simply execute nothing.

To fix this, we instead have to check if the prefix is an actual
suite prefix as opposed to a simple string prefix. We do so by by
inspecting if the first two characters trailing the prefix are
our suite delimiters '::', and only consider the filter as
matching in this case.

git_diff_file: move `id_abbrev`

Move `id_abbrev` to a more reasonable place where it packs more nicely
(before anybody starts using it).

Fix leaks, some warnings and an error

Coverity user model

remote: Handle missing config values when deleting a remote

And give it a default so that some compilers don't (unnecessarily)
complain.

After 1cd65991, we were passing a pointer to an `unsigned long` to
a function that now expected a pointer to a `size_t`.  These types
differ on 64-bit Windows, which means that we trash the stack.

Use `size_t`s in the packbuilder to avoid this.

Compare the preimage to the image; don't compare the preimage to itself.

ci: point to the homebrew libssh2 install