[vxfemboy]

Resolves #171

Summary

This PR eliminates heap allocations, per-shard HashMap lookups, and intermediate vector cloning in FecReassembler::push(), delivering zero-alloc in-place reassembly for clean-path transport and eliminating Vec<Vec<u8>> allocation during Reed-Solomon packet-loss recovery. It also updates the documentation curriculum, architectural specifications, roadmap, and changelog.

Changes

1. Flat Pre-allocated ObjState & Systematic Slicing:
   • Replaced shards: HashMap<u16, Vec<u8>> in ObjState with a 256-bit bitset (seen_mask: [u64; 4]) and pre-allocated contiguous payload: Vec<u8>.
   • Incoming systematic source shards (idx < K) are copied directly into payload[idx * shard_len ..] with zero heap allocation.
   • Clean delivery path (source_count == K) returns the completed object via std::mem::take(&mut state.payload) and truncate(size), completely skipping RS decoding and vector stitching.
2. In-Place RS Recovery (reconstruct_missing_sources):
   • Added decode_source_into and reconstruct_missing_sources in crates/yip-transport/src/rs.rs.
   • Parity reconstruction operates directly on missing shard slices in the pre-allocated payload buffer using a single scratch buffer, eliminating the Vec<Vec<u8>> heap matrix churn.
3. Flat Open-Addressing ObjectTable:
   • Replaced HashMap<u64, ObjState> with ObjectTable, a flat contiguous table using SplitMix64 fast hashing and backward-shift deletion with zero tombstones and zero dynamic heap reallocations during packet ingress/eviction.
   • Pre-allocated FIFO order: VecDeque<u64> ring buffer capacity.
4. Criterion Microbenchmarks:
   • Added benches/fec_reassemble.rs to yip-bench.
5. Documentation, Roadmap & Changelog Sync:
   • Synchronized CHANGELOG.md under ## [Unreleased] with dedicated ### Performance, ### Added, and ### Changed sections.
   • Updated ROADMAP.md moving the zero-allocation hot path (perf(transport): eliminate heap allocations and HashMap churn in FecReassembler hot path #171) and comprehensive docs overhaul (docs: comprehensive README and documentation overhaul for onboarding and clarity #170) to ## Done.
   • Updated README.md, docs/03-architecture.md, and docs/05-benchmarking-and-performance.md with zero-allocation architecture details and Criterion microbenchmark results.

Benchmark Results (Criterion)

Benchmark Scenario	Baseline	Optimized	Speedup
fec_reassemble_clean_k1	361.98 ns	117.11 ns	3.1x faster (-67.6% latency)
fec_reassemble_clean_k3	690.73 ns	173.12 ns	4.0x faster (-74.9% latency)
fec_reassemble_loss_k3_r1	1.0401 µs	631.83 ns	1.65x faster (-39.3% latency)

Verification

• cargo test --workspace: All 387 workspace unit tests + 98 integration tests pass cleanly.
• cargo clippy --workspace --all-targets -- -D warnings: Clean.
• cargo fmt --check: Clean.
• cargo shear: Clean.