Support 128-bit AVX/SIMD comparison logging and mutation

**Is your feature request related to a problem? Please describe.**
When fuzzing high-performance targets, modern codebases frequently use AVX/SIMD instructions for comparisons. Since libafl and libafl_targets cap comparison logging at 64-bit values, the fuzzer is completely blind to 128-bit, 256-bit, and 512-bit vector comparisons significantly limiting mutation effectiveness on these code paths.

**Describe the solution you'd like**
To add native 128-bit comparison support to CmpLog and the mutation framework, with room to extend to 256/512-bit later. The main changes would be: adding a U128 variant to CmpValues, widening CmplogBytes, fixing up the CmpMap dictionary extraction to properly reconstruct u128 halves, and hooking it all into havoc_mutations::dict_insert()

**Describe alternatives you've considered**
CmpValues::Bytes sort of works but loses the semantic aware mutations that make CmpLog actually useful. Manual per-target hooking is another option, but it's messy putting this in the core CmpMap means every CmpLog-enabled target gets the benefit automatically.

**Additional context**
Without this, achieving deep coverage in cryptographic routines, vectorized string parsers, and multimedia libraries remains substantially harder, as magic bytes and equality conditions hidden behind SIMD optimizations go unresolved.