log_parsers: read unittest and Django test-runner output

[k21993]

Summary

• Parse verbose unittest and Django TextTestRunner output in both the generation parser and standalone verifier.
• Detect python -m unittest, manage.py test and Django runtests.py, adding the verbosity each runner needs to expose stable test identities.
• Group subtests under their parent with the worst status winning, including named subtests and nested parameter values, and keep unittest count footers out of the pytest fallback.

Test plan

• uv run --all-extras pytest -q (2151 passed, 6 skipped)
• uv run ruff check .
• uv run ruff format --check .
• Run the focused parser, normalizer and verifier tests (155 passed)
• Generate named-subtest output with the real TextTestRunner and parse it through both parser copies
• Verify the new normalization and stderr tests fail against the pre-fix main branch

Out of scope

• Custom unittest.TestResult output formats.
• Targeting individual unittest files; these commands continue to run their recorded suite.
• Validation stream capture, fixed separately by pr_runtime: give validation's test block stderr, not just stdout #164.

Closes #167

[adithya-s-k]

Useful addition, but a real unittest case is still dropped by both parser copies: with self.subTest("edge case", i=1): self.assertEqual(1, 2).

TextTestRunner prints test_named (pkg.T.test_named) [edge case] (i=1) ... FAIL, plus the matching failure footer. Both parsers return {} because _NAME_RE rejects the message. That loses the failing parent and prevents F2P detection.

Please support named subtests (with and without parameters), retain the parent ID and worst-status behavior, and add tests using real TextTestRunner output in both copies. #164 is now merged, so the stderr dependency is resolved.

[k21993]

Fixed in 34e42c9. The regression test now generates the exact output with a real TextTestRunner and sends it through both parser copies:

• [edge ... case] (i=1) covers a named subtest with parameters and an ellipsis in the message.
• [message only] covers a named subtest without parameters.

Both collapse to pkg.NamedSubTests.test_named, with the existing worst-status behavior. The parser now splits on the final result separator and accepts the full subtest suffix instead of only a parenthesized parameter block.

I also rebased onto current main, so the merged #164 commit is no longer in this PR. Full suite: 2151 passed, 6 skipped; ruff clean.