refactor(math): remove dead/self-referential code from polynomial.rs

[diegokingston]

Remove functions with no production callers (only exercised by their own tests) from crypto/math/src/polynomial.rs:

• compose_fft (was #[cfg(test)], tested only by composition_fft_works)
• pad_with_zero_coefficients + pad_with_zero_coefficients_to_length
• mul_with_ref (no Polynomial Mul operator exists)
• new_monomial

Adjust affected tests: drop self-tests, rewrite the evaluate-coverage tests to build polynomials via Polynomial::new, and inline the single production-test call site (X^8) in prover_tests.

[github-actions]

Codex Code Review

No security, VM, or performance issues found in the PR diff.

One potential compatibility concern: crypto/math/src/polynomial.rs removes public APIs (new_monomial, mul_with_ref, pad_with_zero_coefficients*). Since crypto/math/Cargo.toml defines a publishable library package and these functions were not test-gated, downstream users compiling against them will break. If this crate is intended to have external consumers, consider keeping deprecated wrappers or documenting this as a breaking API change. If it is strictly internal, this is fine.

Tests not run: cargo test was blocked because rustup tried to write under read-only /home/runner/.rustup/tmp.

[claude]

Review: refactor(math): remove dead/self-referential code from polynomial.rs\n\nOverall: Clean removal. No issues found.\n\nRemoves 5 dead functions from polynomial.rs, adjusts downstream tests, and inlines the single non-test use of new_monomial in prover_tests.rs. A grep over the full repo confirms zero remaining callers.\n\nMinor notes:\n\n- X^8 construction verbosity — The 9-element array is correct and the comment helps, but a reader who mis-counts zeros creates a silent bug. Acceptable tradeoff since new_monomial is gone; watch if the pattern proliferates.\n\n- Breaking public API — All removed functions were pub. Semver-breaking for any external consumer. Fine at 0.1.0 with no external callers in this repo — worth a changelog entry if the crate is ever published.\n\n- mul_with_ref algorithm — The removed code contained a valid O(n*m) polynomial multiplication. If a Mul trait impl is added later it will need to be re-derived. Nothing actionable now.\n\nNo security, correctness, or performance concerns.

[MauroToscano]

Why do we even have the Polynomial at this point ? You cannot do much with them, right ?