fix: fingerprint Solidity functions in ctx duplicates (AGE-2)

[saldestechnology]

Summary

ctx duplicates silently skipped all Solidity functions. The MinHash near-duplicate detector normalizes each function into token shingles via tree-sitter, and Solidity has no tree-sitter grammar in this build (it's parsed with solang-parser), so .sol functions never got fingerprints at index time. On a Solidity-heavy repo this made the detector — and the perception of coverage — blind to the largest part of the codebase.

Approach

Rather than add tree-sitter-solidity (which conflicts with the pinned tree-sitter ~0.20.10 — the documented reason the project chose solang in the first place), tokenize Solidity with the solang-parser lexer already in the dependency tree.

• New tokenize_solidity in src/fingerprint.rs: lexes with solang_parser::lexer::Lexer and maps tokens with the same normalization as the tree-sitter path (normalize_leaf): identifiers → ID; string/hex/address/number literals → LIT; keywords and punctuation kept verbatim; comments dropped (the lexer excludes them). Each token's 1-indexed line is derived from its byte offset so file_fingerprints slices by symbol line range exactly as for other languages.
• tokenize short-circuits to this path for Language::Solidity; ts_language still returns None for it.
• Removes the now-false "skipped" surfaces: skipped_languages in duplicates --json no longer lists solidity, the human-readable skip note is gone, and the duplicates/module docs are updated.

Scope note

ctx audit's Duplication score is a separate name-based heuristic that never consults MinHash, so this change fixes ctx duplicates but does not by itself move the audit score. Rewiring audit's duplication category onto fingerprints is a possible follow-up, not included here.

Tests

• New test_tokenize_solidity_normalization (identifiers→ID, literals→LIT, renamed identifiers produce an identical stream).

• Inverted test_solidity_files_produce_no_fingerprints → test_solidity_files_produce_fingerprints.

• Updated tests/duplicates_cli.rs skip-list expectation and added an end-to-end test asserting two structurally-identical .sol functions are detected as a near-duplicate pair.

• cargo fmt --all -- --check ✅

• cargo clippy --all-targets -- -D warnings ✅ (0 warnings)

• cargo test ✅ (304 lib + all integration suites, 0 failures)

Refs AGE-2. Sibling of AGE-4 (modifier edges) and AGE-5 (qualified-call resolution).