docs: document WASM context serialization optimizations and updated benchmarks

[aepfli]

Summary

• Document the three WASM context serialization optimizations in both READMEs: pre-evaluation cache, context key filtering, and index-based evaluation
• Add evaluate_by_index and evaluate_reusable to the main README's exported functions table
• Update update_state response format with new fields (requiredContextKeys, flagIndices, preEvaluated)
• Replace outdated benchmark numbers with current JMH results showing 32x speedup for targeting flags with 1000+ attribute contexts
• Update Java API reference with EvaluationContext overload and UpdateStateResult fields

Key benchmark results (JMH, 1000+ attribute LayeredEvaluationContext)

Scenario	Old json-logic-java	WASM Evaluator	Speedup
Simple flag (cached)	0.023 µs	0.020 µs	~same
Targeting match	409 µs	12.8 µs	32x
Many evals (×1000)	409 µs	11.9 µs	34x

Test plan

• No code changes — documentation only
• Verified all benchmark numbers match actual JMH output

🤖 Generated with Claude Code