This is the moral equivalent of

  git/git@39a87a29ce364ed3337e535adce5973731ba2968

from Adrian Johnson <ajohnson@redneon.com>.

This pulls upstream changes from:

  git/git@8a2e8da367f7175465118510b474ad365161d6b1

  git/git@abf8f9860248d8c213600974742f18dadaa8fbb5

  git/git@407e07f2a6f55e605fda9e90cb622887269f68b5

all by Johannes Sixt <j6t@kdbg.org>.

git_merge_base() returns GIT_ENOTFOUND when it cannot find a merge
base. graph_desdendant_of() returns a boolean value (barring any
errors), so it needs to catch the NOTFOUND return value and convert it
into false, as not merge base means it cannot be a descendant.

Correct grouping of parentheses

git_graph_descendant_of was returning the result of an assignment

Fix bug with multiple iconv conversions in one dir

The internal buffer in the `git_path_iconv_t` structure was not
being reset before the calls to `iconv` were made to convert data,
so if there were multiple decomposed Unicode paths in a single
directory, paths after the first one were being appended to the
first instead of treated as independent data.

Add support for ** matches in ignores

Don't lose our elements when calling git_vector_set()

This is an experimental addition to add ** support to fnmatch
pattern matching in libgit2.  It needs more testing.

Improve submodule cache management

Retry committing locked files on error

I wrote this stuff a while ago and forgot to write tests.  Wanted
to do so now to wrap up the PR and immediately found problems.

The base for the relative urls is determined as follows, with descending
priority:

- remote url of HEAD's remote tracking branch
- remote "origin"
- workdir

This follows git.git behaviour

Const correctness!

Handle an upstream branch for an unborn one

Skip tests on Clar

When the current branch is unborn, git will still mark the current
branch's upstream for-merge if there is an upstream configuration. The
only non-constrived case is cloning from an empty repository which then
gains history. origin's master should be marked for-merge.

In order to do this, we cannot use the high-level wrappers that expect a
reference, as we may not have one. Move over to the internal ones that
expect a reference name, which we do have.

Cloning from an empty repo must set master's upstream to origin's
master, even if neither of them exist.

Fetching from a non-empty origin must then mark the master branch
for-merge. This currently fails.

Fix submodule_is_config_only's return value

No need to find merge base.

Correct a stale reference to GIT_EBAREINDEX

Per @carlosmn, git_index_add is now named git_index_add_bypath.

Skip untracked contained repo contents even with gitlink files

One blame test replies on being run from within the libgit2
repository to leverage having a longer history to play with, but
some bundled versions of libgit2 don't have the whole libgit2
history.  This just skips that test if the repository can't be
opened.

When doing a diff for use in status, we should never show the
content of a git repository contained inside another one.  The
logic to do this was looking for a .git directory and so when a
gitlink plain .git file was used, it was failing to exclude the
directory content.

There was a little bug where the submodule cache thought that the
index date was out of date even when it wasn't that was resulting
in some extra scans of index data even when not needed.

Mostly this commit adds a bunch of new tests including adding and
removing submodules in the index and in the HEAD and seeing if we
can automatically pick them up when refreshing.