0.51.2: Model switching and typing continuity

• In-flight model switching: Telegram model selection can again stop, switch, and continue any interruptible agent run in the current Pi session, including local/TUI work without an active Telegram prompt. The exact model-menu chat/Thread/message supplies fallback continuation ownership, active tools defer abort until settlement, and cancellation/session boundaries clear both selection and target state instead of returning a false busy response.
• Typing continuity: Native presence now refreshes only the assigned Thread every three seconds instead of duplicating each action into aggregate All. The leader admits at most one concurrent action per chat and shares Telegram 429 cooldown across that chat's Thread keys. Compaction preserves a pre-existing agent typing loop instead of claiming and stopping it, addressing both multi-instance flicker and lifecycle gaps.
• Typing status truthfulness: Failed sendChatAction(typing) calls remain bounded structured diagnostics but no longer replace a healthy connected/leader/follower status with error. Optional presence failures therefore cannot misreport transport health after the actual turn and delivery continue successfully.