fix: serialize resource discovery per context

[leno23]

Summary

• Add a per-context async mutex map to K8sClientPool so discovery/refresh calls for the same Kubernetes context cannot race each other.
• Keep discoveries for different contexts parallel by locking only the requested context key.
• Re-check the in-memory registry cache after acquiring the context lock, so a waiting non-force discovery can reuse the fresher registry written by the call that just completed.

Fixes #76

Testing

• cargo fmt --check
• git diff --check
• cargo test discovery_lock_is_per_context -- --nocapture
• cargo check
• Attempted cargo clippy --all-targets --all-features -- -D warnings; it reaches existing unrelated warnings in src/cli/repl.rs (collapsible_match) and src/datafusion_integration/hooks.rs (default_constructed_unit_structs) after this change compiles successfully.

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, add credits to your account and enable them for code reviews in your settings.