feat(ais_core): implement repo_resolve library + minimal migration (#470, Issue #1b)

[SaJaToGu]

Closes #470 (Issue #1b — Wave 1b, Repo-Resolution)

Strict Parallel-Implementation + Comparison-Tests approach

Per the #470 review feedback, this PR follows the
expand-and-contract pattern: implement the new library
functionality in parallel to existing script logic, pin the
existing behaviour with characterization tests, write
comparison tests that show library and legacy agree on the
simple (owner, repo) case, then migrate ONE small
integration point with no broad refactor.

What's in this PR

ais_core/repo_resolve.py — new library

• ResolvedRepo (NamedTuple): owner, repo, remote, local_path
• resolve_from_owner_repo(owner, repo) -> ResolvedRepo
  • No GitHub API call. Validates GitHub naming rules
    (1-39 owner, 1-100 repo, alnum + - _ . + no
    leading/trailing separators), strips whitespace, returns
    ResolvedRepo with canonical
    https://github.com/{owner}/{repo} and local_path=None.
• resolve_from_git_remote(path) -> ResolvedRepo
  • Reads remote.origin.url via local git -C path config --get
    (no network). Parses via urllib.parse for HTTPS, ssh://,
    and git@host:owner/repo forms.
  • Returns ResolvedRepo with local_path set.
• resolve_repo_hint(hint) -> ResolvedRepo
  • Dispatch:
    • existing directory → resolve_from_git_remote
    • "owner/repo" → split + resolve_from_owner_repo
    • bare "repo" → resolve_from_owner_repo with
      GITHUB_USER env-var default

Internal helpers: _validate_component, _read_git_remote_url,
_parse_git_remote_url, _canonical_remote, _parse_http_or_ssh.

tests/test_ais_core/test_repo_resolve.py — 34 tests

• Unit tests for each public function
• URL parser tests (HTTPS with/without .git, with
  trailing slash, SSH short form, ssh:// form, invalid input)
• Characterization test — _legacy_resolve() helper pins the
  pre-0.10.0 Issue #1b — Repo-Resolution extrahieren (ein Script nutzt es) #470 behaviour. NOT exported as a production function;
  test-only, and explicitly documents that the library CHANGES
  whitespace handling (semantic tightening) versus the legacy.
• Comparison tests — library and _legacy_resolve() agree
  on the (owner, repo) case for realistic owners
  (SaJaToGu, python, microsoft).
• No-network test — mock.patch on subprocess.run and
  urllib.request.urlopen confirms resolve_from_owner_repo
  makes no network calls.

scripts/solve_issues.py — ONE small integration point

After the preflight_checks block, this single call exercises the
library code path:

if args.repo:
    from ais_core.repo_resolve import resolve_from_owner_repo
    _resolved_repo = resolve_from_owner_repo(user, args.repo)
    del _resolved_repo  # suppress unused-variable lint

The result is deliberately unused. Per the strict-pattern
directive, the migration is a single integration point that
proves the library works against the script's actual values, with
no broad refactor of the existing args.repo / user handling.
Old logic for "owner from config" stays untouched.

Out of scope (intentional)

• No GitHub API integration — resolve_from_owner_repo does
  not verify the repo exists on github.com.
• No SSH-vs-HTTPS preference — canonical remote is always HTTPS.
• No GitHub Enterprise / sub-group support.
• No routing of args.repo / user through the library beyond
  the one smoke-use call.
• No removal of the test-helper _legacy_resolve() — it is
  pinned in test_legacy_resolve_basic to keep the
  comparison-test infrastructure intact.

Tests

python -m unittest tests.test_ais_core.test_repo_resolve \ tests.test_ais_core.test_json_contract \ tests.test_ais_core.test_run_state \ tests.test_ais_core.test_secret_filter \ tests.test_ais_core.test_issue_resolve

→ 120 tests, 1.068s, all OK.

LOC

File	Lines
ais_core/repo_resolve.py	294
tests/test_ais_core/test_repo_resolve.py	361
scripts/solve_issues.py (delta)	+9

Parent issue

Part of #468 (Release 0.10.0 — AIS Tooling Interface).