Add benchmarks for network, consensus, and LSM subsystems

[michaeljfazio]

Summary

Benchmarks exist for 6 crates (crypto, ledger, mempool, primitives, serialization, storage) but are missing for network throughput, consensus chain selection at depth, LSM compaction, and the node block-apply pipeline.

Current Benchmarks

• dugite-crypto/benches/crypto_bench.rs — Ed25519, VRF, KES operations
• dugite-ledger/benches/utxo_bench.rs — UTxO lookups, inserts, batch operations
• dugite-mempool/benches/mempool_bench.rs — tx admission, removal, conflict detection
• dugite-primitives/benches/hash_bench.rs — hashing operations
• dugite-serialization/benches/serialization_bench.rs — CBOR encode/decode
• dugite-storage/benches/storage_bench.rs — ChainDB read/write

Proposed New Benchmarks

dugite-network

• Protocol message encode/decode throughput
• Multiplexer segment framing overhead
• Pipelined ChainSync message throughput (messages/sec at various pipeline depths)
• Handshake negotiation latency

dugite-consensus

• Chain selection with varying fork depths (10, 100, 1000 blocks)
• VRF leader check throughput (checks/sec)
• Epoch transition processing time
• Header validation throughput

dugite-lsm

• Write throughput (inserts/sec) at various memtable sizes
• Read throughput with varying SSTable counts (read amplification)
• Compaction throughput and write amplification
• Bloom filter false positive rate under load
• Snapshot creation/restoration time

dugite-node (pipeline)

• End-to-end block apply throughput (decode → validate → apply → store)
• Batch block import rate (Mithril replay simulation)
• Ledger snapshot save/restore time at various UTxO set sizes

Infrastructure

All benchmarks should use Criterion for statistical rigor and integrate with the existing nightly CI benchmark workflow (.github/workflows/benchmarks.yml).