Refactor decompress_bmi2_ptr to reduce complexity (#383)

Extracted the match copying logic from `decompress_bmi2_ptr` into a new
helper function `copy_match_bmi2`. This significantly reduces the size
and complexity of the main decompression loop, improving readability
and maintainability. The helper function retains the `unsafe` block
and `#[target_feature]` attributes to ensure correct SIMD code generation.
Removed `#[inline(always)]` from the helper to comply with compiler
stability rules for target features.

Co-authored-by: google-labs-jules[bot] <161369871+google-labs-jules[bot]@users.noreply.github.com>

Author: 404Setup