Add minimal mutant setup

[mbj]

This configures mutant for in top_secret, it intentionally does not:

• Add mutant as a CI gate.
• Attempt to cover any mutation.

This is for after I had a discussion with Luis who invited me to show mutant here ;)

Full mutant run produces this summary:

Mutant environment:
Usage:           opensource
Matcher:         #<Mutant::Matcher::Config subjects: [TopSecret*]>
Integration:     rspec
Jobs:            128
Includes:        ["lib"]
Requires:        ["top_secret"]
Operators:       light
MutationTimeout: 5
Subjects:        61
All-Tests:       116
Available-Tests: 116
Selected-Tests:  116
Tests/Subject:   1.90 avg
Mutations:       1482
Results:         1482
Kills:           1121
Alive:           361
Timeouts:        0
Runtime:         4.91s
Killtime:        101.32s
Efficiency:      2061.66%
Mutations/s:     301.57
Coverage:        75.64%

Note I work on a riddiculous machine (128 threads) will be slower on other computers. And a full pass is only useful for very small projects, regular CI / interactive / agent use would normally go for --since main / --fail-fast.

Claude was so nice to make me this summary on the alive mutations:

  Top alive-mutation hotspots (worth attacking first):

  ┌────────────────────────────────────────┬─────────────────┬──────────────────────────────────────────┐
  │                Subject                 │ Alive (visible) │                   File                   │
  ├────────────────────────────────────────┼─────────────────┼──────────────────────────────────────────┤
  │ Text::Result.with_global_labels        │ 86              │ lib/top_secret/text/result.rb:47         │
  ├────────────────────────────────────────┼─────────────────┼──────────────────────────────────────────┤
  │ Text::Result.from_messages             │ 50              │ lib/top_secret/text/result.rb:34         │
  ├────────────────────────────────────────┼─────────────────┼──────────────────────────────────────────┤
  │ Text::BatchResult.from_messages        │ 48              │ lib/top_secret/text/batch_result.rb:33   │
  ├────────────────────────────────────────┼─────────────────┼──────────────────────────────────────────┤
  │ Text.shared_model                      │ 24              │ lib/top_secret/text.rb:21                │
  ├────────────────────────────────────────┼─────────────────┼──────────────────────────────────────────┤
  │ Text::BatchResult#initialize           │ 5               │ lib/top_secret/text/batch_result.rb:21   │
  ├────────────────────────────────────────┼─────────────────┼──────────────────────────────────────────┤
  │ Text::Result#initialize                │ 3               │ lib/top_secret/text/result.rb:21         │
  ├────────────────────────────────────────┼─────────────────┼──────────────────────────────────────────┤
  │ Text::GlobalMapping#process_result     │ 1+              │ lib/top_secret/text/global_mapping.rb:40 │
  ├────────────────────────────────────────┼─────────────────┼──────────────────────────────────────────┤
  │ Text::GlobalMapping#build_from_results │ 2               │ lib/top_secret/text/global_mapping.rb:26 │
  ├────────────────────────────────────────┼─────────────────┼──────────────────────────────────────────┤
  │ Text#substitute_text                   │ 4               │ lib/top_secret/text.rb                   │
  ├────────────────────────────────────────┼─────────────────┼──────────────────────────────────────────┤
  │ Text.clear_model_cache!                │ 1               │ lib/top_secret/text.rb:38                │
  └────────────────────────────────────────┴─────────────────┴──────────────────────────────────────────┘

  Patterns of weak coverage that jump out:
  - Batch / global-mapping pipeline is the largest gap — most of the 361 alive mutations cluster there. Specs likely
  assert end-to-end shape but not the intermediate transformations (the global-label substitution loop, the
  value_to_label mapping, etc.).
  - Memoization & mutex semantics in Text.shared_model / clear_model_cache! aren't pinned down — the mutex
  synchronization can be removed without breaking any test.
  - Constructor field assignments (@input, @items) — initializers often have alive nil-substitution mutations because
  tests don't read every getter back after construction.
  - Splat usage (*default_filters.keys vs default_filters.keys) — assert_valid_keys happens to accept both shapes, so
  behavior diverges only on edge inputs no test exercises.

  Want me to start with Text::Result.with_global_labels (the biggest single subject), or open a session subject report
  (mutant session subject ...) for the full mutation list on a target?

Most of the alive mutations are "semantically dead code" that can likely just be removed, and many others are a sign of behaviore the code exhibits that probably should not be removed but is "not being pulled into existance by a test that demands that behavior".

I do not expect this to be merged as is will, as I said: Discuss with Luis on this PR ;)

[stevepolitodesign]

@louis-antonopoulos tagging you just so I don't forget.

[github-actions]

This PR is stale because it has been open 30 days with no activity. Remove stale label or comment or this will be closed in 5 days.

[mbj]

@louis-antonopoulos you want to get this merged? I'm happy to rebase.

[louis-antonopoulos]

I still need to evaluate and talk with @stevepolitodesign first!

[mbj]

kk, np at least we have defeated the stale bot for the next 30d ;)