Fix stale diagnostics when cached results exist

[mohsenil85]

Summary

• Fix stale cached diagnostics: When publishDiagnostics results are already cached, getDiagnostics returned them immediately without waiting for rust-analyzer to finish re-analyzing after a file edit. Added a waitForDiagnosticsIdle call in the cached path so in-flight updates settle before returning.
• Increase default idle detection thresholds: maxWaitTime 1s→2s, idleTime 100→200ms — the previous defaults were too short for multi-crate Rust workspaces.
• Increase fallback wait times: Fallback paths use multipliers (4-5x) of the base timeout automatically.
• Make timing configurable: New optional diagnostics field in cclsp.json:

  {
    "diagnostics": {
      "maxWaitMs": 2000,
      "idleMs": 200
    }
  }

  Both fields are optional and default to 2000ms / 200ms respectively.

Context

In a multi-crate Rust workspace (~5 crates), calling get_diagnostics immediately after an edit consistently returned stale results. The root cause was the cached diagnostics fast path (the if (cachedDiagnostics !== undefined) branch), which returned instantly. The fallback paths with their waits were never reached because diagnostics from a previous analysis were already cached.

Test plan

• Tested manually in a 5-crate Rust workspace: introduced a syntax error, called get_diagnostics, confirmed it reported the error
• Fixed the error, called get_diagnostics again, confirmed the error was gone
• Previously both calls returned stale results from the prior state

🤖 Generated with Claude Code