Performance benchmark

We should have a benchmark that simulates typical language server operations (e.g. open a workspace, open some files, edit, rename, navigate, etc). It will serve as an integration test, a performance health check, and as a way to evaluate potential performance optimizations.

See also:

• Deno LS blog post: https://deno.com/blog/optimizing-our-lsp
  • benchmark: https://github.com/denoland/deno/blob/03a76f33ad0808c659116721a757958ad054dd25/cli/bench/lsp.rs#L77-L141
  • flamegraphs to identify performance bottlenecks
• rust-analyzer syntax: https://github.com/rust-lang/rust-analyzer/blob/master/docs/dev/syntax.md
  • alternate syntax representations could reduce memory usage, but need to test
• https://code.visualstudio.com/api/language-extensions/language-server-extension-guide#testing-the-language-server

Start by reviewing the Deno LSP benchmark linked in the issue, the rust-analyzer syntax notes, and the VS Code language-server testing guide. Define representative workspace, file, edit, rename, and navigation operations, then establish how the benchmark will serve as an integration test and performance health check. Done means the benchmark runs reproducibly and can expose performance changes.