fix(crypto): add safety check for odd lengths in vector_karatsuba

[kimyee1]

Rationale

The vector_karatsuba function is only used for Falcon signature generation, which always operates on power-of-two lengths (e.g., 512). However, the function is accessible, and calling it with odd lengths (like 9 or 17) currently causes a panic due to out-of-bounds indexing or silent math errors.

Following the maintainer's recommendation in #3550 to "make the power-of-two precondition explicit," this PR adds a safety check to fail fast with a clear error message.

What I changed

• Added assert!(n.is_power_of_two()) at the entry of vector_karatsuba.
• This prevents silent math errors or confusing index panics for unsupported lengths.
• Added a regression test karatsuba_panics_on_non_power_of_two to prove the guard works.

Test plan

• Ran cargo test -p miden-crypto locally.
• Verified that the new test correctly catches the odd-length case and successfully panics.

Fixes #3550

[github-actions]

Automated check (CONTRIBUTING.md)

Findings:

• Link an issue in the PR body (e.g., "Fixes Implement SHA256 in Miden Assembly #123").

Next steps:

• Link a relevant issue (e.g., "Fixes Implement SHA256 in Miden Assembly #123") and ensure it is assigned to you.
• See CONTRIBUTING.md for expectations.
• If this is a false positive, comment: /quality-review.

[kimyee1]

All checks are now green after applying the nightly rustfmt formatting. 🟢

Quick recap: vector_karatsuba previously assumed power-of-two lengths, so odd-length inputs could panic or silently corrupt results. This PR makes the precondition explicit with an assertion, and adds a #[should_panic] regression test to prove the guard catches it.

Happy to adjust the error message or test placement if you prefer a different shape.

cc @huitseeker

[Al-Kindi-0]

We should close in favor of #3629, which resolves #3487 with a complete documentation, testing, and hardening pass over the Falcon math module, covering the change proposed here.