x86: Explore performance improvements for XKCP version of Keccak-F1600

[hanno-becker]

As noted by @manastasova, our x86 intrinsics code for Keccak-F1600 from XKCP has optimization potential. We should investigate and potentially update.

See #1484 for performance data.

[manastasova]

The code in #1484 is based on the original mlkem-native/XKCP implementation (which are currently identical) but introduces several new optimizations that are not present in the upstream XKCP code. These optimizations account for the performance improvements rather than mlkem-native's FIPS 202 implementation being out of sync with XKCP.

[hanno-becker]

Thank you @manastasova, I clarified the issue.