When we read the header, we want to know the size and type of the
object. We're currently inflating the full delta in order to read the
first few bytes. This can mean hundreds of kB needlessly inflated for
large objects.

Instead use a packfile stream to read just enough so we can read the two
varints in the header and avoid inflating most of the delta.

Annotated commits: differentiate between the ref names and the description

Differentiate between the ref_name used to create an annotated_commit
(that can subsequently be used to look up the reference) and the
description that we resolved this with (which _cannot_ be looked up).

The description is used for things like reflogs (and may be a ref name,
and ID something that we revparsed to get here), while the ref name must
actually be a reference name, and is used for things like rebase to
return to the initial branch.

Test that we can properly abort a rebase when it is initialized by a
revspec.  This ensures that we do not conflate revspecs and refnames.

Remove traces of `git_blob_create_fromchunks`

Allow creating copies of `git_reference` objects.

Rebase improvements with IDs

tag: ignore extra header fields

While no extra header fields are defined for tags, git accepts them by
ignoring them and continuing the search for the message. There are a few
tags like this in the wild which git parses just fine, so we should do
the same.

⚡ some warnings

When rebasing with IDs, we do not return to the `branch`,
we remain in a detached HEAD state.

When `init`ing a rebase from a detached HEAD, be sure to remember
that we were in a detached HEAD state so that we can correctly
`abort` the object that we just created.

Instead of `open`ing a rebase and `abort`ing that, test that we can
`abort` a rebase that has just begun with `init`.

Proxy configuration

We were downloading the jar from within an block which only runs for
MSVC. Move the download to the start of the test so it gets downloaded
for both.

Improve star-star matching

Instead of threading the state down to the larger loop, let's have the
loop where we detect the double star so each of them are easier to read.

In order to match the star-star, we disable the flag that's looking for
a single path element, but that leads to searching for the pattern in
the middle of elements in the input string.

Mark when we're handing a star-star so we jump over the elements in our
attempt to match the part of the pattern that comes after the star-star.

While here, tighten up the check so we don't allow invalid rules
through.

Running clar directly on appveyor makes it think the command returned
failure, so it stops the tests. Running it via ctest lets it go through.

We leave this up to the scheme in the url field. The type should only
tell us about whether we want a proxy and whether we want to auto-detect
it.

It takes a bit for the propxy to get ready to accept connections, so
start it before the build so we can be reasonably sure that it's going
to be ready in time.

I don't quite recall what we do in the other places where we use this,
but we should pass this payload to the callbacks.

The path is not something that you use for proxies, so make use of the
new optionality of the path when extracting URL parts.

When we're dealing with proxy addresses, we only want a hostname and
port, and the user would not provide a path, so make it optional so we
can use this same function to parse git as well as proxy URLs.

It is currently unused; it will go into the remote's options.

Add missing ')' to callbacks documentation

Super minor, but it was bugging me.

There was a missing closing paren in the docs.