fix(rust_brain): preserve HLC on snapshot restore and gossip receive

[cursor]

Bug and impact

After v0.6.0 added HLC monotonic enforcement, restore_from_file() and gossip.receive() still assigned fresh HLC timestamps instead of preserving/restoring them from snapshots or gossip payloads.

Concrete trigger:

1. Node writes remember("k", "v1", hlc=(5000, 10, "nodeA")) and snapshots to disk.
2. After crash/restart, restore_from_file() assigns a fresh HLC (wall, 0, uuid) instead of (5000, 10, "nodeA").
3. A causal successor write remember("k", "v2", hlc=(5000, 11, "nodeA")) raises TimestampRegression → silent data loss.
4. Gossip without HLC can overwrite fresher local data; gossip with stale HLC is not rejected.

Root cause

restore_from_file() constructed MemoryNode without the hlc field (defaulting to _hlc.now()), and gossip.receive() called remember() without passing hlc/ts_ns. bulk_write() had the same gap.

Fix

• Add _parse_hlc() for snapshot/wire normalization with legacy ts_ns fallback
• Restore HLC and advance global clock (_hlc.update()) on snapshot restore
• Pass HLC through bulk_write and gossip.receive() with stale-update rejection
• Hold lock during restore_from_file() and gc_expired() to prevent races

Validation

• pytest tests/test_enterprise_backup.py tests/test_gossip.py tests/test_rust_brain.py tests/test_rust_brain_concurrency.py — 37 passed
• New tests: test_restore_preserves_hlc_for_replication, gossip HLC acceptance/rejection tests