perf(discovery): pre-parse and cache whenExpr in Collector during Pack validation

[blue4209211]

Description

⚡ Bolt Performance Optimization: Pre-parse and cache whenExpr in Collector during Pack validation

💡 Problem

During discovery_inventory runs across server fleets (up to 5,000 targets per request), runInventory calls pack.Select(res.Facts) for every single host.
Inside Pack.Select(), for every collector with a when guard, parseWhen(c.When) was previously called repeatedly for every host. parseWhen performs multiple strings.Split, string allocations, map lookups, and slice allocations for AST structures ([][]comparison) — even though Pack.validate() already checked the syntax when the pack was loaded.

For a realistic content pack with ~12 guarded collectors across a 1,000-host fleet, this caused 12,000 redundant AST parsing cycles and ~75,000 heap allocations on the inventory sweep path.

⚡ Solution

1. Added an unexported expr *whenExpr field to Collector.
2. In Pack.validate(), stored the compiled *whenExpr into c.expr during pack validation/verification.
3. In Pack.Select(), evaluated c.expr directly (falling back to parseWhen(c.When) if c.expr is nil for defensive backward compatibility with ad-hoc unvalidated structs).
4. Pre-allocated the run slice in Select() with make([]Collector, 0, len(p.Collectors)).
5. Added benchmark coverage (BenchmarkPackSelect and BenchmarkExamplePackSelect).

📊 Benchmark Results (Apple M2 Pro)

# Before:
BenchmarkExamplePackSelect-10       331514      3690 ns/op      3536 B/op      75 allocs/op
BenchmarkPackSelect-10             1772901       671.8 ns/op     464 B/op      14 allocs/op

# After:
BenchmarkExamplePackSelect-10      2942226       399.2 ns/op    1024 B/op       1 allocs/op
BenchmarkPackSelect-10            14205189        98.47 ns/op    176 B/op       1 allocs/op

• Speedup: 9.24x faster (89.2% latency reduction) for 17-collector production packs; 6.82x faster for 3-collector packs.
• Memory: 71.0% reduction in bytes allocated per host selection.
• Allocations: 98.7% reduction in allocations (from 75 allocs down to 1 single alloc per host).

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Enhancement (non-breaking change which improves existing functionality)
• Refactor (non-breaking change which improves code structure)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation
• CI/CD

How Has This Been Tested?

• Unit tests
• Manual testing

Ran make lint (0 issues), go test -race ./... (all packages pass), and go test -bench="BenchmarkPackSelect|BenchmarkExamplePackSelect" -benchmem ./pkg/proxy/discovery.

Checklist

• CLA signed (the CLA bot will prompt on your first PR)
• make validate passes (fmt + lint + test)
• Docs updated if the wire shape, config surface, or proxy module behavior changed

[gemini-code-assist]

Code Review

This pull request introduces performance optimizations to the discovery pack parsing and evaluation logic. Key changes include replacing static map lookups with a switch statement to eliminate heap allocations, pre-parsing and caching guard expressions (when expressions) during pack validation to avoid repeated parsing during host sweeps, and optimizing signature verification to avoid duplicate line parsing. Additionally, several benchmarks have been added to track these performance improvements. There are no review comments, so no feedback is provided.