Fix GitLab clone URLs and authentication in pr_diff Dockerfiles

[abhinavgautam01]

Fixes #93

Summary

GitLab merge requests were mined through the correct provider, but their emitted pr_diff Dockerfiles cloned a hardcoded GitHub URL. This could fail the environment build or target a different repository.

Build clone URLs from the source repository URL, preserving the host and full path. Normalize supported SSH-style inputs to HTTPS and select GITHUB_TOKEN or GITLAB_TOKEN using the existing authentication helper.

Quote token expansion, remove credentials supplied in source URLs and retain the clean origin reset after cloning. Update documentation to describe provider-specific authentication.

Validation

Add 15 regression cases covering GitHub compatibility, GitLab URLs, nested paths, SSH-style inputs, credential stripping and public/authenticated clone commands.

The shell tests execute the generated clone commands with a fake Git executable and dummy tokens, checking argument handling and origin cleanup without network access.

• Python 3.12, 3.13 and 3.14: 731 passed, 5 skipped on each.
• Ruff lint, formatting and whitespace checks passed.
• Source distribution and wheel builds passed.
• Clean wheel installation with locked dependencies, CLI version check and dependency compatibility check passed.

Scope

Private GitLab MR diff fetching remains unsupported under #65. This change fixes emitted clone URLs and build-time authentication; it does not establish end-to-end private GitLab mining support.

[KNambiarDJsc]

I reviewed and tested 719238d:

• CI-equivalent run on Linux: ruff is clean, and the full suite passes on py3.12 and py3.14 (730 passed).
• Generated Dockerfiles, main vs this PR:
  • --repo pallets/click: the only change is the quoted token expansion, x-access-token:"${GITHUB_TOKEN}"@github.com/.... The shell treats it the same, so GitHub tasks are unaffected.
  • --repo https://gitlab.com/python-devs/importlib_resources: main emitted ARG GITHUB_TOKEN and cloned https://github.com/python-devs/importlib_resources.git. With this PR it emits ARG GITLAB_TOKEN, clones through oauth2:"${GITLAB_TOKEN}"@gitlab.com/..., and uses the correct clean URL.
• The quoting tests are thorough. They run the real RUN body through /bin/sh with a fake git, including a token containing spaces, *, $HOME and $(false).

LGTM. Two non-blocking follow-ups:

1. docs/reference/AUTH.md is still GitHub-only. Its "Private repos at task build time" section says the Dockerfile declares ARG GITHUB_TOKEN= and clones through x-access-token:<token>@github.com/..., so a consumer following it for a GitLab task would pass the wrong build arg. Mentioning GITLAB_TOKEN and oauth2: there would match the updated pr_diff.md.
2. Windows test runs. The 6 test_clone_shell_uses_correct_token_and_scrubs_origin cases fail on Windows with FileNotFoundError, because there's no /bin/sh. CI is Linux-only, so there's no CI impact. Adding pytest.mark.skipif(not Path("/bin/sh").exists(), reason="needs a POSIX shell") would keep the suite usable there; #108 tracks the other Windows issues.

[adithya-s-k]

thanks @abhinavgautam01, merged! appreciate the GitLab clone and auth fixes, especially the coverage around credentials.