test(corekit): drop the dead conventions.rst candidate and the stale #912 tense in _sentinel_section

[JarryShaw]

_sentinel_section() in tests/corekit/test_sentinel_exports_unit.py still carries two leftovers from #912's move of conventions.rst, both surfaced by the cross-review of #919.

1. A dead first candidate. Lines 135-136 on 3766c3c09:

for candidate in ('docs/source/conventions.rst',
                  'docs/source/contributing/conventions.rst'):

docs/source/conventions.rst has not existed since #912 landed as 9806f16aa, so the first iteration's path.is_file() is always false. The loop still works — it falls through to the real path — but it is dead code that reads as though both locations were live.

2. A stale tense in the docstring. Lines 129-131 describe the move as in progress:

the move to :file:docs/source/contributing/conventions.rst that GitHub pull request #912 is making

#912 has merged. #919 corrected the :file: path on that same docstring's first line but left the tense and the dead branch beneath it.

Neither is a behaviour defect: the helper resolves the right file today, and the else: raise AssertionError fallback is correct. It is hygiene, and the reason to fix it is that a reader cannot tell the dead branch from a deliberate compatibility shim.

Blocked on #911. That issue's work has this exact file dirty in flight, editing the module docstring and the SENTINELS table. A second edit to it now would either conflict or silently revert. Unblocks when #911 merges. #919 is also still open against this file, so the clean start point is after both.

[JarryShaw]

Unblocked — #911 merged as c7342cac1. The file contention is gone: tests/corekit/test_sentinel_exports_unit.py is no longer held by an open PR.

Both leftovers this issue describes survived the merge unchanged — the dead first candidate 'docs/source/conventions.rst' in _sentinel_section()`'s fallback tuple, and the docstring still describing #912's move as in progress. #922 deliberately left the tuple alone on my instruction, and flagged it rather than editing it.

Removing blocked.

[JarryShaw]

Fixed and pushed directly to main as 5e0ec3889 — test-only, which is authorised here, so it did not need a PR.

Both leftovers gone:

• _sentinel_section() no longer walks a candidate tuple. It resolves docs/source/contributing/conventions.rst directly, since the first entry docs/source/conventions.rst has not existed since docs: move the process pages out of the docs top level (#901) #912 landed as 9806f16aa — the loop always fell through, so it was dead code that read as though both locations were live.
• The docstring no longer says docs: move the process pages out of the docs top level (#901) #912 "is making" the move. It records that the move landed, and why the section-marker slicing (rather than line numbers) is what survived it.

Comment-only and docstring-only; no assertion changed. Verified: tests/corekit/test_sentinel_exports_unit.py 18 tests OK, tests/project/test_isort_clean.py OK, and grep finds zero occurrences of the pre-move path anywhere under pcapkit/, tests/ or docs/ — so this closes the last of #912's trailing references rather than just the two named here.

Closing.