Rollup merge of #145963 - heiher:src-analysis-lsx, r=lqd

Add LSX accelerated implementation for source file analysis

This patch introduces an LSX-optimized version of `analyze_source_file` for the `loongarch64` target. Similar to existing SSE2 implementation for x86, this version:

- Processes 16-byte chunks at a time using LSX vector intrinsics.
- Quickly identifies newlines in ASCII-only chunks.
- Falls back to the generic implementation when multi-byte UTF-8 characters are detected or in the tail portion.

Author: Zalathar