Don't canonicalize symlink targets; our win32 path canonicalization
routines expect an absolute path.  In particular, using the path
canonicalization routines for symlink targets (introduced in commit
7d55bee6d, "win32: fix relative symlinks pointing into dirs",
2020-01-10).

Now, use the utf8 -> utf16 relative path handling functions, so that
paths like "../foo" will be translated to "..\foo".

Add a function that takes a (possibly) relative UTF-8 path and emits a
UTF-16 path with forward slashes translated to backslashes.  If the
given path is, in fact, absolute, it will be translated to absolute path
handling rules.

The path canonicalization functions on win32 are intended to
canonicalize absolute paths; those with prefixes.  In other words,
things start with drive letters (`C:\`), share names (`\\server\share`),
or other prefixes (`\\?\`).

This function removes leading `..` that occur after the prefix but
before the directory/file portion (eg, turning `C:\..\..\..\foo` into
`C:\foo`).  This translation is not appropriate for local paths.

On Windows platforms, we need some logic to emulate symlink(3P) defined
by POSIX. As unprivileged symlinks on Windows are a rather new feature,
our current implementation is comparatively new and still has some
rough edges in special cases.

One such case is relative symlinks. While relative symlinks to files in
the same directory work as expected, libgit2 currently fails to create
reltaive symlinks pointing into other directories. This is due to the
fact that we forgot to translate the Unix-style target path to
Windows-style. Most importantly, we are currently not converting
directory separators from "/" to "\".

Fix the issue by calling `git_win32_path_canonicalize` on the target.
Add a test that verifies our ability to create such relative links
across directories.

On Windows platforms, we automatically clean up the thread-local storage
upon detaching a thread via `DllMain()`. The thing is that this happens
for every thread of applications that link against the libgit2 DLL, even
those that don't have anything to do with libgit2 itself. As a result,
we cannot assume that these unsuspecting threads make use of our
`git_libgit2_init()` and `git_libgit2_shutdow()` reference counting,
which may lead to racy situations:

    Thread 1                    Thread 2

    git_libgit2_shutdown()
                                DllMain(DETACH_THREAD)
                                git__free_tls_data()
    git_atomic_dec() == 0
    git__free_tls_data()
    TlsFree(_tls_index)
                                TlsGetValue(_tls_index)

Due to the second thread never having executed `git_libgit2_init()`, the
first thread will clean up TLS data and as a result also free the
`_tls_index` variable. When detaching the second thread, we
unconditionally access the now-free'd `_tls_index` variable, which is
obviously not going to work out well.

Fix the issue by converting the code to use fiber-local storage instead
of thread-local storage. While FLS will behave the exact same as TLS if
no fibers are in use, it does allow us to specify a destructor similar
to the one that is accepted by pthread_key_create(3P). Like this, we do
not have to manually free indices anymore, but will let the FLS handle
calling the destructor. This allows us to get rid of `DllMain()`
completely, as we only used it to keep track of when threads were
exiting and results in an overall simplification of TLS cleanup.

Prefer `off64_t` internally.

Prefer `off64_t` to `git_off_t` for internal visibility.

64 bit types are always 64 bit.

When creating a symlink in Windows, one needs to tell Windows whether
the symlink should be a file or directory symlink. To determine which
flag to pass, we call `GetFileAttributesW` on the target file to see
whether it is a directory and then pass the flag accordingly. The
problem though is if create a symlink with a relative target path, then
we will check that relative path while not necessarily being inside of
the working directory where the symlink is to be created. Thus, getting
its attributes will either fail or return attributes of the wrong
target.

Fix this by resolving the target path relative to the directory in which
the symlink is to be created.

When deleting a symlink on Windows, then the way to delete it depends on
whether it is a directory symlink or a file symlink. In the first case,
we need to use `DeleteFile`, in the second `RemoveDirectory`. Right now,
`p_unlink` will only ever try to use `DeleteFile`, though, and thus fail
to remove directory symlinks. This mismatches how unlink(3P) is expected
to behave, though, as it shall remove any symlink disregarding whether
it is a file or directory symlink.

In order to correctly unlink a symlink, we thus need to check what kind
of file this is. If we were to first query file attributes of every file
upon calling `p_unlink`, then this would penalize the common case
though. Instead, we can try to first delete the file with `DeleteFile`
and only if the error returned is `ERROR_ACCESS_DENIED` will we query
file attributes and determine whether it is a directory symlink to use
`RemoveDirectory` instead.