Use UTF-16 positions in the markup language server

[narmi924]

Summary

• map incoming LSP UTF-16 positions to UTF-8 byte offsets for completion and hover
• map parser byte columns back to UTF-16 diagnostic ranges
• clamp CRLF line ends and malformed or out-of-range positions safely
• advertise UTF-16 position encoding and cover CJK, emoji, and protocol round trips

Why

The markup parser tracks byte columns, while LSP clients use UTF-16 code units by default. The server previously treated protocol character values as byte offsets and emitted parser byte columns directly, so CJK text or emoji earlier on a line shifted completion, hover, and diagnostic locations.

Validation

• direct markup LSP suite: all 10 tests passed
• zig build test-markup-cli
• zig build validate
• scripts/gate.sh fast origin/main: validation and all frontend, native, and mobile example groups passed on this branch; root tests reached only the existing Windows CRLF source-pattern failures fixed by Make build validation portable on Windows #157
• detached combined validation with Make build validation portable on Windows #157: root tests, validation, and all frontend/native/mobile example groups passed

[vercel]

@narmi924 is attempting to deploy a commit to the Vercel Labs Team on Vercel.

A member of the Team first needs to authorize it.