pygit2: resolve_rev raise a proper exception with ~<n> like referen…

[karajan1001]

fix: #30
Current pygit2 backend's resolve_rev didn't raise a proper exception when dealing with
HEAD~n like references.

1. Catch the exception in resolve_rev and reraise it into a proper format.
2. Add a test for it.

In resolve_refish both HEAD~n like references and non-exist references raise a ValueError, can't distinguish them. And I don't think we should let Inner exception InvalidSpecError be thrown outside of scmrepo. But this method is a relatively slow one, because we need to iter over all of the remotes.

Closes treeverse/dvc#7434

[pmrowla]

I'm still confused as to why this PR is needed, since exp show -n ... and pygit resolve_rev(f"HEAD~{n}") worked prior to the recent refactors. It seems like this might need some more investigation into why resolve_rev started failing here (since the exception was not raised at all before)

Is the issue that the exception is raised when n is invalid (and that relative commit does not actually exist)? Or that it is still raised when HEAD~n points to a valid commit (and that the problem is with using ~ at all)?

[karajan1001]

I'm still confused as to why this PR is needed, since exp show -n ... and pygit resolve_rev(f"HEAD~{n}") worked prior to the recent refactors. It seems like this might need some more investigation into why resolve_rev started failing here (since the exception was not raised at all before)

Is the issue that the exception is raised when n is invalid (and that relative commit does not actually exist)? Or that it is still raised when HEAD~n points to a valid commit (and that the problem is with using ~ at all)?

It raises when n is invalid, I add two test cases here.

    assert git.resolve_rev("HEAD~1") == init_rev # this is valid , and works pretty fine in previous.

    with pytest.raises(RevError):
        git.resolve_rev("HEAD~3") # in this case, the previous version will raise `InvalidSpecError` instead of `SCMError`

In the previous version of DVC, if we input an invalid number of n, it will raise an unexcepted error.

$ dvc exp show -n 100 --no-pager
ERROR: unexpected error - refs/remotes/origin/HEAD~11: the given reference name 'refs/remotes/origin/HEAD~11' is not valid

Having any troubles? Hit us up at https://dvc.org/support, we are always happy to help!

[pmrowla]

@karajan1001 this needs a rebase for the CI fixes