feat: local Flair instance on branch VMs with background sync from rockit

[tps-anvil]

Problem

Branch VMs (e.g. tps-anvil) currently rely on a flair-proxy over the branch wire connection to reach Flair on rockit:9926. This is fragile:

• Internet/tunnel outages break memory access entirely
• Adds latency to every memory operation
• The HTTP relay (MSG_HTTP_REQUEST/MSG_HTTP_RESPONSE) isn't fully wired up

Solution

Run a local Flair instance on each branch VM, with background sync from rockit (authoritative instance).

• Local Flair on tps-anvil at localhost:9926 (or 8787)
• Background sync process: pulls new/updated memories from rockit periodically
• OpenClaw memory-flair plugin points to local instance
• Memory writes go local-first, sync to rockit in background
• Survives outages — local reads/writes always work

Scope

1. Install and configure Flair on branch VMs (launchd/supervisor, no systemd)
2. Build/configure a sync script or daemon (bidirectional or rockit→branch+branch→rockit)
3. Update OpenClaw plugin config to use local URL
4. Document setup so it's repeatable for new VMs

Notes

• flair-sync.mjs exists but syncs files→Flair, not instance→instance
• Observatory sync is Flair→Observatory push, not peer sync
• No new npm deps; prefer bun scripts or extend existing Flair tooling