Show one look per Bot on every home in face-to-face rooms

[Birfy]

Fixes #182

What

• Owner's look reaches the other home. A Bot with no stored shape or color is drawn at home with an id-picked look (core characters.js), but Bot Social got nothing, so the other home invented a handle-seeded one. syncBotLook now sends the look the home draws (shownLook in plugins/social/src/look.js, the same functions as core, checked equal by a test). A stored image or emoji is sent as is, in the order BotMark draws them (image, emoji, shape).
• Look before join. A new (room-only) identity gets its look before registerOnce returns, so the first members[] the peer sees already has it.
• My own Bot in the waiting dialog uses its roster look. Before, it was handle-seeded because the Bot list was read before the join created the identity (the purple acorn in the 01:00 self-test). It is now found by its room record (remote: false, local id) or by this join's joinedAs.
• Chat bubbles in a face room draw a friend's Bot with its seat's look (members[]), the same as Members, the room avatar and the waiting tiles (the teal blob vs blue whale).
• People show their initial on a color picked from their name, the same on both homes, not a grey silhouette (PersonMark({ name })).

Scope. Room titles were made consistent across homes in #186. The number in the title and the Past rows are #304, and own-bubble avatars are #457, so neither is in this PR. Two Bots that both carry the owner-stored default Chief look (whale) still look the same: each home draws its owner's own look, and the owner name tells them apart.

Test

• ARCH_DOCS_BASE=origin/main TZ=UTC npm run verify: pass, 0 failures (core 1019, social 279, tree 80, timeline 80, scripts, interop, live-e2e unit).
• New tests: the generated look equals core for 200+ ids, and the sent look equals what core draws. A Bot with no look gets it on Bot Social before POST /face-rooms/join. The waiting dialog uses the roster look right after a join and after a poll. remoteSpeaker uses the seat look in a face room. PersonMark draws the initial and a stable color.
• Not run: a live two-home face-room check on Mac instances.

Risk tier: T3 (estimate, matching the risk-tier report check; the controller sets the label)

[Birfy]

Self-review: PASS
Tip SHA: 75da40b
Ran:

• ARCH_DOCS_BASE=origin/main TZ=UTC npm run verify: exit 0. Core lint + arch-docs OK (20 files), plugin lint, build-client check, ds-bot-patches check, core 1019/0 fail, social 279/0, tree 80/0, timeline 80/0, scripts, interop and live-e2e unit tests all pass.
• node scripts/ds-bot-patches.mjs generate then check: core equals ds-bot plus patches.
• CI at this tip: tests-ok green (test, plugins social/tree/timeline, package). Risk tier (report only): T3.
• Diff read against the owner design rules:
  • The look goes out only through MCP update_bot_profile, as in A friend's Bot ignores the owner's chosen avatar: the same Bot looks different in each home (follow-up to #84) #137. There is no publish, grant or approval change.
  • Untrusted peer fields (members[] color and avatar) are only read through the existing sanitized participant record.
  • UI copy stays in English.
    Not run:
• A live two-home face-to-face check on the Mac instances (3080/3081) or on fresh A/B homes. The next self-test should re-run the repro in this issue: waiting dialog, tiles, chat, Members and sidebar on both homes, then a second room.