fix(cli): parse markdown command frontmatter on Windows CRLF/BOM

[zy6p]

Summary

• normalize markdown command content before frontmatter parsing by stripping UTF-8 BOM and converting CRLF to LF
• make frontmatter regex robust for empty frontmatter and optional trailing newline
• add regression tests for CRLF and BOM parsing scenarios
• add loader-level regression test to ensure markdown command descriptions are read correctly from BOM+CRLF files

Testing

• npm run test --workspace=packages/cli -- src/services/markdown-command-parser.test.ts src/services/FileCommandLoader-markdown.test.ts

Fixes #2053

[Aayushyaash]

@zy6p

This looks like the same class of Windows newline/BOM parsing issue.

There is a related open PR (#1890) that introduces a normalizeContent() helper which strips UTF-8 BOM and normalizes CRLF/CR → LF before parsing. The goal there is to centralize newline normalization so different parsers don’t each implement their own handling.

If it makes sense, this markdown frontmatter parsing could potentially reuse that helper instead of adding another normalization step in the CLI parser. That would keep newline/BOM handling consistent across the repo and avoid duplicated logic.

What do you think?

[zy6p]

@Aayushyaash Thanks for the suggestion. I updated this PR to reuse the shared normalizeContent() helper from core utils (instead of parser-local normalization), so BOM/newline handling stays centralized.\n\nAlso added coverage for:\n- core normalizeContent() (BOM, CRLF, CR-only, mixed line endings)\n- CLI markdown frontmatter parsing with CRLF/BOM/CR-only content\n\nLatest commit: 05f2224.

[tanzhenxin]

LGTM!