fix(mutmut): make scripts/ test-loader module names match mutmut's path-derived keys

[cdeust]

Summary

Fixes #262: mutation testing could not attribute mutants for any
scripts/ module because tests_py/scripts/*.py tests loaded their
subjects under synthetic or bare module names
(_cortex_launcher_deps, generate_pip_constraints, bare
import wiki_backfill_ids, ...), while mutmut keys mutant trampolines
on the path-derived dotted name (scripts.launcher_deps,
scripts.generate_pip_constraints, ...) it computes purely from each
mutated file's path relative to the repo root. Every scoped run aborted
early: "Stopping early... tests recorded trampoline hits but none match
any mutant key."

Root cause traced by reading mutmut's own source
(mutmut/mutation/trampoline.py's record_trampoline_hit(f"{orig_func.__module__}...")
vs mutmut/utils/format_utils.py::get_mutant_name, which derives the key
from the file path alone). __module__ is bound to whatever name
importlib.util.spec_from_file_location/module_from_spec gave the
loaded module — a test-only artifact completely disconnected from the
path-derived key mutmut expects.

Precedent already existed in this repo: PR #236 fixed the identical
defect for test_check_doc_claims.py alone and filed the survivor
backlog it surfaced as #235, rather than folding a much larger
test-writing effort into the attribution-fix PR. This PR applies the
same fix to every other affected file (test_check_doc_claims.py
excluded — already fixed, and owned by the concurrent #235 worktree),
and follows the same split for the backlog it surfaces: filed as #263.

Changes (3 commits)

1. fix(mutmut) — rename every affected test's module-loader handle
   to the dotted scripts.<file> name mutmut expects (14 files); widen
   [tool.mutmut] also_copy (uv.lock, requirements/, pyproject.toml)
   so generate_pip_constraints.py's test — which reads all three
   directly — mutates cleanly instead of every mutant raising
   FileNotFoundError.
2. test — new regression test
   (test_launcher_deps_stays_bare_importable_by_launcher_py) proving
   AC 2: launcher_deps.py stays loadable by a bare import at real
   runtime (a genuine python3 scripts/launcher.py subprocess, not a
   test-loader path). Verified it actually catches a regression
   (temporarily broke the import, watched it fail, reverted).
3. docs — resync the advertised test count across README.md,
   CONTRIBUTING.md, CLAUDE.md, docs/ASSURANCE-CASE.md,
   .bestpractices.json, and the regenerated assets/badge-tests.svg.
   Rebased onto main after it independently merged json_native.py: 14 surviving mutants under the repository's own committed mutmut scope #250/pyright gate resolves a different tree-sitter-language-pack than the documented dev install #253/normalize_date_to_iso silently re-anchors an unresolvable timezone abbreviation to UTC #252
   (baseline moved 6439 -> 6549 while this branch was in flight);
   resolved the six suite-size files' conflicts on their merits —
   main's structure, count recomputed from a live collection on the
   rebased tree (6550 = main's 6549 + this PR's 1 new test), not a
   delta applied to either side's stale number.

Verification (acceptance criteria from #262)

1. Real mutant tallies now produced for both AC-named modules (previously
   an early abort with zero usable result):

   scripts/mutation_check.sh tests_py/scripts/test_launcher_deps.py scripts/launcher_deps.py
     -> 213 mutants, 128 killed, 85 survived, 0 no-tests
   
   scripts/mutation_check.sh tests_py/scripts/test_generate_pip_constraints.py scripts/generate_pip_constraints.py
     -> 243 mutants, 95 killed, 65 survived, 83 no-tests
   

2. launcher_deps.py bare-importability asserted, not assumed (new
   regression test above).
3. also_copy widened with uv.lock/requirements//pyproject.toml.
4. Survivors triaged per §12.4: filed as Mutation-test survivors in scripts/launcher_deps.py and scripts/generate_pip_constraints.py (85 + 65), now that #262 makes them measurable #263 (85 + 65, matching the
   #236 -> #235 precedent) — this PR fixes the attribution seam, not
   the pre-existing coverage gaps it now makes measurable for the first
   time.

Local, on the rebased tree (head 16d296c), quoted for reviewer sanity:
pytest tests_py/ -> 6550 passed, 117 subtests passed in 161s.
pytest tests_py/scripts/ -> 301 passed, 117 subtests passed.
ruff check + ruff format --check on every touched .py file: clean.
scripts/check_doc_claims.py --test-count 6550 -> OK.
scripts/generate_repo_badges.py --check --test-count 6550 -> OK (5
checked). scripts/check_marketplace_pins.py -> OK.

Completion Ledger

Path introduced by this diff	Asserting test / evidence
Renamed loader in 14 tests_py/scripts/*.py files	tests_py/scripts/ full run: 301 passed before AND after the rename commit (0 test-code behavior change)
Widened [tool.mutmut] also_copy	generate_pip_constraints.py scoped mutation run produces a real 95/65/83 tally instead of FileNotFoundError aborts
New test_launcher_deps_stays_bare_importable_by_launcher_py	Self-verifying: shown to fail against a deliberately broken launcher.py import, then pass against the real one
Doc/badge count resync (post-rebase: 6550)	check_doc_claims.py --test-count 6550 + generate_repo_badges.py --check --test-count 6550 both green; full suite collects exactly 6550

Boy-scout (§14)

• ruff format flagged one line-length issue introduced by my own edit in
  test_setup_py_backend_skip.py — fixed in the same commit (formatter
  auto-fix, verified ruff format --check clean afterward).
• Rebase reconciliation (this session): resolved 6 conflicting doc/badge
  files by recomputing the count from a live pytest --collect-only -q
  on the rebased tree rather than taking either side's stale number, and
  corrected the reconciliation commit's own message/date to describe
  what actually happened (was previously drafted against the pre-rebase
  6440 number before the conflict was known).
• No other defects seen in touched material this session.

Deferred (§14.3 — outside this PR's blast radius, filed with acceptance criteria)

• Mutation-test survivors in scripts/launcher_deps.py and scripts/generate_pip_constraints.py (85 + 65), now that #262 makes them measurable #263 — the 85 + 65 mutation survivors these two modules now expose
  (plus 83 uncovered "no tests" mutants in generate_pip_constraints.py).
  Triaging/killing them is a test-writing effort against orchestration
  and subprocess-heavy code, independent of the attribution-seam fix this
  PR makes.

🤖 Generated with Claude Code