This commit introduces AVX2-vectorized assembly implementations of the
Number Theoretic Transform (NTT) operations used in ML-DSA (FIPS 204).
These optimizations improve performance of ML-DSA key generation,
signing, and verification operations on x86_64 platforms
with AVX2 support.

The implementation adds the following functions:
- ml_dsa_poly_ntt_avx2: Forward NTT transformation
- ml_dsa_poly_ntt_inverse_avx2: Inverse NTT transformation
- ml_dsa_poly_ntt_mult_avx2: NTT-domain polynomial multiplication

Key implementation details:
- Uses YMM registers to process 8 32-bit coefficients in parallel
- Employs Montgomery reduction for modular arithmetic
- Implements NTT butterfly operations across multiple transform levels
- Includes dedicated zeta table for INTT to reduce cycles
- Runtime capability check via ml_dsa_ntt_avx2_capable() using
  OPENSSL_ia32cap_P to detect AVX2 support

The C code in ml_dsa_ntt.c is updated to dispatch to AVX2 implementations
at runtime when available, with automatic fallback to the portable C
implementation on platforms without AVX2 support.

Build system changes:
- Added GENERATE rule for ml_dsa_ntt-x86_64.s from Perl assembly
- Conditional assembly inclusion based on target architecture
- Works with both libcrypto and FIPS provider builds

Signed-off-by: Marcel Cornu <marcel.d.cornu@intel.com>
Signed-off-by: Tomasz Kantecki <tomasz.kantecki@intel.com>

Signed-off-by: Marcel Cornu <marcel.d.cornu@intel.com>