Code-quality pass: deduplicate move-making/primitives and tighten lints

[kirillbobyrev]

Summary

A quality-focused pass over the chess core and the tooling config. No behaviour changes — purely deduplication, modern idioms, and stricter lints. Verified against the full perft suite (including the shakmaty differential test), the incremental-hash invariant test, the UCI integration tests, and a self-play smoke game.

Refactoring (refactor: commit)

• make_move: the Zobrist hash-update boilerplate (self.hash ^= generated::get_piece_key(Piece { player, kind }, square)) was repeated ~12 times across four sub-functions, each of which re-scanned the piece bitboards to find the mover. Introduced toggle_piece/move_piece helpers that keep the hash in sync, and determine the moving piece once via Pieces::at, then dispatch. Shorter and does less work per move (the old code ran every move through all four sub-functions).
• Piece ↔ char: replaced the two 12-arm char↔Piece matches with a single source of truth, PieceKind::to_char/from_char; Piece now derives Copy/Clone/Eq.
• Castling generation: encapsulated the static castle geometry in a const Castle table and iterate over it, deleting the 9-argument try_castle helper and its parallel 9-tuple destructure.
• UCI: the go command now carries a Go parameter struct instead of an 8-field enum variant feeding an 8-argument Engine::go; extracted Engine::limits. Both "too many arguments" clippy smells are gone, and the parser tests shrank via ..Go::default().
• Completed the Bitboard operator set with toggle + BitXorAssign.

Harness (chore: commit)

You mentioned the lint config hadn't been revisited in a while — I measured the churn of every commented-out lint and turned on the ones that are already clean:

• rustc lints: enabled unreachable_pub, trivial_numeric_casts, unused_qualifications, unused_import_braces, unused_lifetimes, unused_extern_crates, macro_use_extern_crate, absolute_paths_not_starting_with_crate (all +0 warnings today). Left missing_docs / let_underscore_drop / unused_results deferred with a note — they still fire on undocumented public API and the search stubs.
• clippy groups: enabled correctness (deny), plus complexity and style (warn), alongside the existing perf/suspicious denies. pedantic (+844), nursery (+26) and cargo stay off — dominated by literal-separator noise and transitive-dependency version reports, not real issues.
• rustfmt: adopted style_edition = "2024".
• Fixed the one real style finding this surfaced (match → ? in the FEN parser).

Testing

• cargo test: 130 tests + doctests pass.
• cargo test --release -- --ignored: all 12 expensive perft / shakmaty-differential tests pass (re-run after the castling and make/move refactors).
• Self-play smoke game (120 plies, incl. castling) over real UCI — no crashes, legal throughout.
• cargo +nightly fmt --check clean; cargo clippy --all-targets --all-features clean of denied lints.

🤖 Generated with Claude Code

https://claude.ai/code/session_01GVrKvkMe4xDmkRAksacBzk

---

Generated by Claude Code