test(benchmark): add criterion microbenchmarks and e2e benchmark CLI tool

[slow2342]

Summary

• Add Criterion microbenchmarks for storage engine, Raft log store, and codec serialization
• Add aether-bench CLI tool for end-to-end gRPC benchmark testing
• Add benchmark compile check to CI pipeline

Changes

• benches/storage.rs — RocksDB StorageEngine benchmarks: put/get/delete/batch_write/scan with varying data sizes
• benches/raft_store.rs — Raft log store benchmarks: append_entries/entries_read/save_hard_state/compact
• benches/codec.rs — Serialization benchmarks: rkyv, MVCC key encoding, KeyIndex encode/decode
• src/bin/aether-bench/ — End-to-end benchmark CLI tool:
  • put — concurrent write benchmark with round-robin worker distribution
  • range — concurrent read benchmark with data pre-fill
  • report — latency percentiles (P10-P99.9), RPS, wall time, error distribution
• Cargo.toml — Add criterion dev-dependency, [[bench]] targets, [[bin]] target, default-run
• .github/workflows/ci.yml — Add cargo bench -- --test compile check

Benchmark results (single node, release mode)

Scenario	RPS	P50	P99
PUT 1 client	4,057	0.24 ms	0.33 ms
PUT 10 clients	14,139	0.68 ms	1.10 ms
RANGE 10 clients	31,701	0.31 ms	0.52 ms

How to verify

cargo bench -- --test                    # Verify criterion benchmarks compile
cargo build --bin aether-bench           # Verify CLI tool compiles
cargo bench                              # Run all criterion benchmarks
cargo run --bin aether-bench -- put --total 1000 --val-size 256  # Run e2e benchmark