Interactive ego-network (force-directed) graph viz on profiles #3

What

Replace the static follow-dot grid on a profile with an interactive ego-network force-directed graph: the identity + its follows, with the follow-edges among those follows drawn so clusters/communities show.

Feasibility (probed — it's ready)

• Data is fast: fetch the ego's follow-list, then one $in batch over the alters' follows docs (~20–50ms), intersect in memory.
• Density is real for active accounts: one ego (390 follows) yields 18,428 intra-network edges; nolim1t's followers show 557 edges in a 50-sample. Inactive/niche accounts are genuinely sparse (expected social-graph structure, not a bug).

Backend

GET /api/graph/:pubkey → { nodes, edges }:

• nodes = the ego + up to ~120 of its follows (cap by the follow's own follower-count so dense egos stay legible), each with { pubkey, name, picture, followers }.
• edges = follow-relationships among the selected nodes (optionally thresholded for very dense egos).

Frontend

• Hand-rolled canvas force simulation (~120 lines, zero new client deps — matches beacon's from-scratch style). Nodes rendered as the gradient avatars, sized by follower-count; faint edges; drag + click-a-node to navigate to /<hex>.
• Degrade gracefully: sparse egos render a light radial; dense egos render a capped, readable cluster.

Acceptance

• Profile shows an interactive graph; dense egos stay smooth via the node cap; clicking a node routes to that identity.
• Quality improves automatically as the raw-WS indexer (Run beacon's raw-WS indexer in production and retire the legacy firehose #2) backfills more follow-lists.