Improve git client tests - #14351

Description

Layers 1-3 replaced broad git.Client mocks with real local Git repositories. This follow-up makes that converted suite smaller and clearer without changing production code.

It replaces Command.Output tests that compiled temporary Go executables with the existing helper-process seam, removes an incidental URL data variation, narrows redundant remote setup, splits tables that mixed success and failure strategies, and gives parser/value tests explicit Given/When/Then structure. Distinct transport outcomes, parser equivalence classes, credential scoping and clearing, --, --ff-only, --set-upstream, and command modifier contracts remain covered.

How did you test this change?

Given the rewritten command error scenario, when I temporarily stopped Command.Output from preserving the process exit status, I saw the focused test fail with an expected status of 128 and an actual status of 0. I reverted the mutation.

Given a repository with no remote resolution metadata, when I temporarily removed the special handling for Git's no-matching-config status, I saw the focused remotes test fail with exit status 1. I reverted the mutation.

Given a git+ssh URL, when I temporarily removed that supported transport from URL classification, I saw the specific classifier scenario fail. I reverted the mutation.

Key points

• The diff removes 67 net test lines and leaves statement coverage unchanged at 86.1%.
• Package timing varied enough between runs that I am not claiming a wall-clock improvement, but TestOutput no longer performs two nested go build operations.
• The tests remain in package git because parser and command-construction contracts are deliberately internal; moving them would require exporting production internals.
• No production code changed.

Notes for reviewers

Start with git/command_test.go for the feedback-time change, then review the split parser scenarios in git/url_test.go and the reduced setup and table shape in git/client_test.go.

Parent draft: #14350.

Authorship and follow-up

Who wrote this:

• A human wrote it.
• An agent wrote it under close human direction.
• An agent wrote it independently, and no human has guided the implementation beyond the initial prompt.

Who answers review comments:

• @williammartin will read and reply directly. Name the account.
• An agent will draft replies and @username will read them before they are posted.
• Nobody has explicitly committed to replying.