delete_head and Head.delete historically accept either Head objects
or a str name of a head.  Adjust the typing to match.  This
unfortunately requires suppressing type warnings in the signature of
RemoteReference.delete, since it inherits from Head but does not
accept str (since it needs access to the richer data of
RemoteReference).

Using assignment to make add an alias for create unfortunately
confuses mypy, since it loses track of the fact that it's a
classmethod and starts treating it like a staticmethod.  Replace
with a stub wrapper instead.

Turns out aec58a9 did the right thing for /packed/ refs, but didn't work
correctly on /unpacked/ refs.  So this patch gives unpacked refs the
same treatment.

Without the fix here, the test added will cause this traceback:

======================================================================
ERROR: Check that we find .git as a worktree file and find the worktree
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/pjones/devel/github.com/GitPython/git/test/lib/helper.py", line 92, in wrapper
    return func(self, path)
  File "/home/pjones/devel/github.com/GitPython/git/test/test_repo.py", line 938, in test_git_work_tree_dotgit
    self.assertIsInstance(repo.heads['aaaaaaaa'], Head)
  File "/home/pjones/devel/github.com/GitPython/git/util.py", line 893, in __getitem__
    raise IndexError("No item found with id %r" % (self._prefix + index))
IndexError: No item found with id 'aaaaaaaa'

Woops.

Things I've learned:
- test_remote doesn't work currently if you start on a branch.  I think
  it never did?
- Because of 346424da, all *sorts* of stuff in the test suite doesn't
  work if you name your development branch "packed-refs"
  (This seems like a bug...)

Signed-off-by: Peter Jones <pjones@redhat.com>

Fixes #7

[ci skip]

More to come, especially when it's about strings

W291 trailing whitespace