feat(hyperswarm): add negentropy sync with sqlite store

[Bushstar]

This PR introduces Negentropy-based anti-entropy sync for the Hyperswarm mediator to replace connect-time full-history sync between upgraded nodes. The goal is to keep peer sync fast and bounded as operation history grows, while remaining backward compatible with legacy peers.

Why this change

• Previous behavior shared large/full operation history on peer connect, which increasingly blocked mediator throughput.
• Negentropy lets peers exchange compact set differences (have/need) instead of bulk history.

What was added

• Negentropy protocol flow in Hyperswarm mediator:
  • new message flow for reconciliation and transfer (neg_open, neg_msg, neg_close, ops_req, ops_push)
  • per-peer sync session state (mode, rounds, pending ids, activity)
• Backward compatibility:
  • capability negotiation in ping
  • upgraded↔upgraded uses negentropy
  • upgraded↔legacy falls back to existing legacy sync path
• Persistent sync store:
  • SQLite-backed ordered store for operation sync index
  • in-memory store for tests
  • mediator persists only operations accepted by gatekeeper
• Bootstrap on upgrade/startup:
  • if sync-store is empty, one-time backfill from gatekeeper exportBatch
  • startup depends on successful bootstrap to avoid partial sync state
  • wipe and bootstrap again if op store has >1% drift from gatekeeper
• Gatekeeper contract extension:
  • importBatch result now includes rejectedIndices
  • mediator filters index-rejected items before persistence pipeline
  • accepted-only persistence is finalized using gatekeeper processing results

Performance safeguards

• Removed full DB export for upgraded peers (legacy only).
• Added adapter freshness controls to avoid repeated rebuild work:
  • dirty flag + built-at age check
  • single in-flight rebuild guard
  • background prebuild when store changes and no active session
• Kept sync concurrency conservative (single active negentropy session).