Add server performance benchmarks and CI regression checks

[ViTeXFTW]

Summary

• Add synthetic and real-corpus server latency benchmarks covering startup, diagnostics, completion, semantic tokens, and edit bursts.
• Add Criterion analysis benchmarks for workspace indexing, reference completion, and ranged semantic tokens.
• Add CI workflows and threshold-based performance reporting for pull requests and corpus runs.
• Remove the obsolete VS Code preview enable setting.

Tests

• Added automated benchmark execution in CI.
• Added scripts/check-performance.py for regression thresholds and absolute latency ceilings.
• Relevant local test commands were documented in CONTRIBUTING.md.

Checklist

• I kept the change focused.
• I ran the relevant checks or explained why they were skipped.
• I did not commit generated artifacts, game source, corpus data, or secrets.