iOS UX: make active-agent indicators open channel-scoped tool activity #5066

Summary

On iOS, active-agent indicators should be reliable navigation targets into that agent's existing channel-scoped tool activity. Today the activity transcript exists, but access is inconsistent and unnecessarily hidden.

Current behavior

• The channel member row becomes directly tappable only while isWorking is true.
• When the agent is not currently marked working, View activity is still available only through the trailing overflow menu.
• Other active/typing indicators do not provide a consistent, obvious path to the same tool transcript.
• A presence reconnect mismatch can remove or contradict the state users rely on to discover the entry point.

Expected behavior

1. Tapping an agent's visible typing/working indicator should open that agent's activity for the current channel.
2. Agent rows and profiles should expose a consistent View activity action regardless of transient working state.
3. The destination should preserve channel context and show the existing live tool-call transcript and connection/error state.
4. If activity cannot be loaded after reconnect, show the actual observer connection state rather than silently disabling navigation.

Source references

• members_sheet.dart only assigns the row tap callback when the bot is currently in the typing-derived working set.
• members_sheet.dart keeps View activity in the overflow action sheet.
• agent_activity_sheet.dart already accepts both channelId and agentPubkey and displays the live transcript.
• observer_subscription.dart already filters observer frames by agent and channel.

Scope

This is an iOS discoverability and navigation UX issue, not a request to invent a new tool-call transport or transcript. It is a focused companion to the broader mobile activity proposals in #3907 and #4169.

Related THREAD/client visibility issue: #5061.