Submodule init, update, caching, reload, status

[jorio]

This PR fleshes out submodules a bit.

• Submodule.init & update: Allow initing and updating Submodule instances when you already have them instead of having to go through Repository.init_submodules and update_submodules (which would do a redundant submodule lookup). Those two functions still exist in Repository, but they are now convenience wrappers around the new init/update functions in all the submodules in the repo.
• Repository.submodule_cache_all & submodule_cache_clear: Expose libgit2's submodule caching features. Speeds up submodule lookups which can be very slow in some repos.
• Submodule.reload: Reload cached info in a Submodule instance if you have reason to believe it has changed externally. Removes the need to do a new submodule lookup to get fresh information.
• Repository.submodule_status: Get the status of a submodule (is it uninitialized, does its workdir have any changes, did the head move, etc.).

---

Notes about API design: The PR brings in new enums (GIT_SUBMODULE_STATUS_...) that I encapsulated as enum classes (IntFlag, IntEnum) in enums.py.

This allows for friendlier development – function signatures can say they return a SubmoduleStatus instead of just an int, and calling repr on a SubmoduleStatus returns more descriptive text than just a number, such as:
<SubmoduleStatus.IN_HEAD|IN_INDEX|IN_CONFIG|WD_UNINITIALIZED: 135>

This isn't how pygit2 currently exposes enums to Python code – but if you think it'd improve pygit2, we could progressively convert the existing enums to IntFlag/IntEnum classes. (The raw GIT_* values would still be available for legacy code.)

If you'd rather not have enums.py for consistency with the existing API, let me know and I'll edit this PR accordingly.

[jdavid]

Merged, thanks.

My only concern is the API consistency of the Repository class:

add_submodule
init_submodules
listall_submodules
lookup_submodule
update_submodules
submodule_status (NEW)
submodule_cache_all (NEW)
submodule_cache_clear (NEW)

I think it would be best to do like the remotes and have:

.submodules.add
.submodules.init
.submodules.__iter__
.submodules.__getitem__
.submodules.update
.submodules.status
.submodules.cache_all
.submodules.cache_clear

With the old function names kept for backwards compatibility, as deprecated.

What do you think? Do you mind to make a new PR in that direction?

[jorio]

I agree, this looks better. I'll submit a new PR this week!