refactor: zero OID handling

[weihanglo]

• Refactor zero Oid initialization in preparation for SHA256 support (libgit2 add a new type field to git_oid struct)
• Reflect that upstream git_oid_iszero is deprecated. Use git_oid_is_zero instead.

(Spun off from #1202)

[ehuss]

Thanks!

I have some doubts if it makes sense to allow kind to be 0 in all situations, since that seems to be an invalid kind. But I don't immediately see any concerns.

[weihanglo]

Thanks!

I have some doubts if it makes sense to allow kind to be 0 in all situations, since that seems to be an invalid kind. But I don't immediately see any concerns.

Good call out!

For output param specifically, kind being 0 should be fine. libgit2 seems to always set out->type based on the type parameter passed to the function or the surrounding context.

For input param, user should not use Oid::zero. It guess somehow we should deprecate it in high level API during sha256 stabilization. Even with zero-initialized oid, libgit2 tends to call git_oid_is_zero and git_oid_is_type_valid to validate input param. I didn't do a full thorough check but seems likely fine.