feat(cli): add gitt miner languages to rank file types by scoring weight and method

[glorydavid03023]

Summary

Language weight is only half of what decides a changed file's worth — the other half is how it is scored. Tree-sitter languages are scored on semantic AST tokens (high ceiling), plain extensions score per added line, and documentation/config extensions are additionally capped at MAX_LINES_SCORED_FOR_NON_CODE_EXT. Today a miner has no way to see this split, so it isn't obvious that (e.g.) editing Rust is worth far more per change than editing a capped Markdown file.

This adds gitt miner languages — a read-only command that loads the same programming_languages.json the validator scores against and prints extensions ranked by weight, each tagged with how it is scored and its tree-sitter language:

• Scoring method per extension: token (tree-sitter AST), line-count, or line-count-capped (doc/config).
• --code-only shows only the token-scored extensions (where AST/token score actually applies).
• --top N, --search <substr> (case-insensitive), and --json-output, matching the existing gitt miner conventions.
• Footer/JSON summary: counts by method and the non-code line cap. No network call or PAT required.

Example:

$ gitt miner languages --code-only --top 8
     Scoring Weights by File Extension
             (tree-sitter only)
┌───┬─────┬────────┬─────────┬─────────────┐
│ # │ Ext │ Weight │ Scoring │ Tree-sitter │
├───┼─────┼────────┼─────────┼─────────────┤
│ 1 │ c   │  2.000 │  token  │ c           │
│ 2 │ cc  │  2.000 │  token  │ cpp         │
│ 3 │ cpp │  2.000 │  token  │ cpp         │
│ … │ …   │  …     │  …      │ …           │
└───┴─────┴────────┴─────────┴─────────────┘

8 of 135 extensions shown | 101 token, 11 line-count, 23 capped doc/config

The scoring-method classification reuses the canonical NON_CODE_EXTENSIONS / MAX_LINES_SCORED_FOR_NON_CODE_EXT constants (the same ones tree_sitter_scoring uses) so it cannot drift from the scorer. Display/aggregation logic is split into small pure helpers (_scoring_method, _filter_languages, _rank_languages, _summarize_languages, _build_language_rows) that are independently unit-tested.

Related Issues

None — net-new miner UX. Complements the repository-weight view in a separate PR; this one is about file-type weight and scoring method.

Type of Change

• Bug fix
• New feature
• Refactor
• Documentation
• Other (describe below)

Testing

• Tests added/updated — tests/cli/test_miner_languages.py (method-classification unit tests incl. the non-code-overrides-tree-sitter edge case, plus end-to-end command tests: table, JSON, --code-only, --top, no-match, and empty-weights error path)
• Manually tested — ran against the real programming_languages.json (135 extensions; 101 token / 11 line-count / 23 capped); verified ranking, method tags, and all flags.

Checklist

• Code follows project style guidelines (ruff check + ruff format clean)
• Self-review completed
• Changes are documented (command help text + miner group docstring updated)