Refactor NTT butterfly iterations & remove redundant input arrays and padding

[doran2728]

Summary

Targeted performance and structural optimizations to the core NTT module.

Changes

Redundant Array Cleanup (Domain.lean, Forward.lean)

• Removed inputArray, zeroPad, and pad helper functions
• The forward pipeline now evaluates directly on raw polynomials without an intermediate array copy

Butterfly Iteration Refactor (Forward.lean, Inverse.lean)

• Rewrote the Cooley-Tukey butterflyStage outer loop in both the forward and inverse transforms
• Before: for base in [0:D.n] with an inner if base % blockSize == 0 guard
• After: for block in [0:D.n / blockSize] with base := block * blockSize
• This direct block iteration eliminates redundant modulo checks, yields faster wall-clock execution, and significantly simplifies the upcoming formal correctness proofs

[erdkocak]

thanks for the effort, since tests pass and the changes were necessary I am merging