test: 136 surviving mutants in the four path-guard modules (pre-existing, outside the changed lines)

[cdeust]

Wiring Python mutation testing (scripts/mutation_check.sh, mutmut 3.x) while fixing the py/path-injection alerts surfaced a pre-existing backlog. only_mutate takes file globs, not line ranges, so a scoped run on a pre-existing file mutates the whole file — separating what the change introduced from what was already there.

The blocking tier is clean. Every mutant on a line the path-injection fix touched is killed: 0 survivors across all six files (classified by matching each mutant's removed source text against the diff's changed lines).

The wide sweep is not, in the three pre-existing files the fix touched:

Module	Test selection	Survivors (none on changed lines)
cortex_viz/infrastructure/wiki_read.py	tests/test_wiki_read.py	67
cortex_viz/server/http_standalone_static.py	tests/test_static_path_traversal.py	53
cortex_viz/server/http_file_diff.py	tests/test_git_diff_engine.py, tests/test_file_diff.py	16

The two new modules are at 0 survivors (shared/path_containment.py 21/21 killed, infrastructure/file_sandbox.py 11/11), and server/git_diff_engine.py is at 0 survivors for the whole file.

This is coding-standards §12.5's non-blocking critical-zone sweep, deferred under §14.3 rather than absorbed silently: the survivors are pre-existing gaps in code the path-injection PR did not change, and closing 136 of them is its own body of work.

One concrete example worth leading with, because it is in a security guard: in serve_static, mutating not safe_name or safe_name.startswith(".") to not safe_name and safe_name.startswith(".") survives — no test distinguishes the two, so the empty-name and dot-prefixed-name arms of that filter are not independently pinned.

Reproduction

./scripts/mutation_check.sh tests/test_wiki_read.py cortex_viz/infrastructure/wiki_read.py
./scripts/mutation_check.sh tests/test_static_path_traversal.py cortex_viz/server/http_standalone_static.py
./scripts/mutation_check.sh tests/test_git_diff_engine.py,tests/test_file_diff.py cortex_viz/server/http_file_diff.py

Acceptance criteria

• Each of the three modules reaches 0 surviving non-equivalent mutants, one module per iteration (§12.5 ratchet).
• Every survivor is either killed by a new test asserting the observable behaviour it pins, or recorded as a provable equivalent mutant with a written rationale — no third option.
• Survivors that turn out to mark dead code are removed rather than tested (§9); the containment primitive's own rstrip was exactly this case and was deleted, not covered.
• Run notes recorded alongside the existing JS ones (tests/js/MUTATION_NOTES.md has the precedent) so the numbers are reproducible.