While we download the remote's remote-tracking branches, we don't
download the tag. This points to the tag auto-follow rules interfering
with the refspec.

Fix duplicate basenames to support older VS on master

With Visual Studio versions 2008 and older they ignore the full path to files and only check
the basename of the file to find a collision. Additionally, having duplicate basenames can break
other build tools like GYP.

This fixes https://github.com/libgit2/libgit2/issues/3356

Error handling: use buffers, improved OOM handling

When an error state is an OOM, make sure that we treat is specially
and do not try to free it.

Allow restoring a previously captured oom error, by
detecting when the captured message pointer points to the
static oom error message.  This means there is no need
to strdup the message in giterr_detach.

Error messages that are detached are assumed to be dynamically
allocated.  Passing a pointer to the static oom error message
can cause an attempt to free the static buffer later.  This change
checks if the oom error message is about to be detached and detaches
a copy instead.

 filebuf: remove lockfile upon rename errors

index: stage an unregistered submodule as well

Handle ssh:// and git:// urls containing a '~' character.

For such a path '/~/...' the leading '/' is stripped so the server will
get a path starting with '~' and correctly handle it.

We previously added logic to `_add_bypath()` to update a submodule. Go
further and stage the submodule even if it's not registered to behave
like git.

When we pass the path of a repository to `_bypath()`, we should behave
like git and stage it as a `_COMMIT` regardless of whether it is
registered a a submodule.

Resolve documentation warnings

Remove extra semicolon outside of a function

Without this change, compiling with gcc and pedantic generates warning:
ISO C does not allow extra ‘;’ outside of a function.

iterator: skip over errors in diriter init

case-insensitive check for WWW-Authenticate header

Fixes issue #3338

Instead of allocating a brand new buffer for each error string we want
to store, we can use a per-thread buffer to store the error string and
re-use the underlying storage. We already use the buffer to format the
string, so this mostly makes that more direct.

We don't want the iterator to make us stop whenever we hit an unreadable
dir. We should instead move over to the next item.

An error here will typically mean that the directory was removed between
the time we iterated the parent and the time we wanted to visit it in
which case we should ignore it.

Other kinds of errors such as permissions (or transient errors) also
better dealt with by pretending we didn't see it.

When we have an error renaming the lockfile, we need to make sure
that we remove it upon cleanup. For this, we need to keep track of
whether we opened the file and whether the rename succeeded.

If we did create the lockfile but the rename did not succeed, we
remove the lockfile. This won't protect against all errors, but
the most common ones (target file is open) does get handled.

When we fail to rename, we currently leave the lockfile laying
around. This shows that behaviour.

Normalize submodule urls before looking at them

Allow adding a submodule through git_index_add_bypath

refdb: delete a ref's reflog upon deletion

filter: make sure to close the stream even on error

Make libgit2 work on Windows Vista again

Increment `git__n_inits` before doing `init_once`.