cleanRemote: sanitize peer-controlled names (strip @ and control chars), not just length-cap #75

Part of #9 (display names). Surfaced by the PR #74 review (finding 1, LOW).

Where

core/src/host/hooks.js cleanRemote builds the local record for a friend's published Bot:

name: String(bot.name ?? remote.handle).slice(0, 60)

It length-caps the display name but does not strip @, newlines, or control characters. homeName and owner are peer-controlled the same way.

Why it matters

• After Fix: a remote Bot's @handle mention renders as the raw handle, not a Bot mention (#71) #74 the client rewrites known handles to these display names in user-facing UI (mention chips, sidebar previews via nameMentions), so a peer name that itself contains a handle-like @… token renders looking like a live mention of another Bot. The client side is now single-pass (displays the name verbatim, no rescan), but the confusing text still comes from the record.
• Peer names can also reach model-visible text (e.g. delivery fromName at hooks.js, and room logs) — the Security: untrusted-content fence escape and unsanitized remote display names (PR #2 review) #9 lesson says sanitize remote strings (strip controls/newlines, cap length) and fence them where they stay model-visible. Worth tracing those paths once the record is clean.

Suggested direction (for discussion, not prescriptive)

• Sanitize at the record boundary in cleanRemote: strip @ and control characters/newlines from name, homeName, owner, keep the 60-char cap.
• Negative tests: a published name containing another member's handle; a name with newlines/control characters.