fix(but): resolve bare paths for staged files in commit -p (#13318)

[sidfishus]

🧢 Changes

Fix bare-path resolution in IdMap so but commit -p foo.txt works for staged files.
Add unit + integration tests for bare path (positive) and zz: prefix (negative).

☕️ Reasoning

non_colon_uncommitted_filename: I thought creating a separate test for cases where there's no colon / prefix makes more sense instead of adding them to the existing colon_uncommitted_filename. That's the reproducer for the bare-path case (#13318).

colon_uncommitted_filename: I only added zz:assigned → empty here, a guard so zz: stays scoped to the unassigned bucket and doesn't regress.

commit.rs: commit_with_bare_path_on_staged_file is the reproducer for that style of test. I also added commit_with_zz_prefix_on_staged_file_fails as the matching negative case. I didn't add comments because the tests are self explanatory and this is consistent with the other tests.

Issue discussion: Per #13318, fix the reported case where a single staged foo.txt wasn't found. Qualified IDs (oo, i0:foo.txt) still work. Duplicate paths across areas still hit the existing ambiguity error; no new disambiguation logic in this PR.

This seems like a suitable issue to pick up for my first contribution.

AI Notice: I have peer programmed this with an AI agent. I've spent a lot of time reproducing and understanding the problem, and considering how best to name the functions (I got confused between uncommitted files and uncommitted/unassigned (zz)). This is what I've finalised on, but happy to be led by somebody with more experience of the codebase.

🎫 Affected issues

Fixes: #13318

[Byron]

Thanks a lot for giving it a shot!

I think this is an area where @jonathantanmy2 would have the most insight.

[sidfishus]

To help with review and give proof here's the 2 new automated tests failing without the changes:

And the end to end run failing without the changes:

And then the 2 new tests succeeding with fix:

And the end to end run succeeding with the fix:

[Byron]

Thanks a lot @sidfishus, this helps for sure!

Let me also reel in @davidpdrsn who has been working no a new implementation of but commit, to see if this fix is already included there as well.