feat: add concrete profiling guide for Kelora optimization

[dloss]

Add comprehensive profiling guide in dev/profiling-guide.md with actionable
optimization strategies for improving Kelora's performance.

Key content:

• Quick start commands for macOS (samply/Instruments) and Linux (flamegraph)
• 6 known Kelora hotspots with optimization opportunities:
  • JSON parsing (30-50% runtime): simd-json, custom deserializer
  • Regex compilation (5-20%): caching, literal optimization
  • Rhai execution (10-40%): native filters, AST precompilation
  • String allocations (5-15%): Cow, string interning
  • Parallel overhead (0-30%): batch tuning, work-stealing
  • Format detection (1-5%): caching, heuristics
• 10 concrete optimization ideas ranked by impact (high >20%, medium 5-15%, low <5%)
• 6-step performance improvement workflow (baseline → profile → identify → optimize → verify → test)
• 3 practical before/after code examples
• Architecture-specific profiling guides (JSON, Rhai, parallel, parsers)
• Common pitfalls and troubleshooting
• Flamegraph interpretation with example diagnoses
• CI integration example

This guide provides concrete, actionable advice for improving Kelora's
performance, not just profiling tool usage. Referenced from AGENTS.md
for developer discoverability.