#21346: [AI-assisted] Telegram: add reaction state machine with fallback and cleanup

## Summary This PR adds a Telegram-side status reaction state machine with compatibility fallbacks and cleanup on completion. For Telegram message processing, the bot now transitions reaction states through: - queued - thinking - tool execution - done / error It also clears the reaction after response completion, so reactions do not linger on user messages. ## Why In Telegram DM/group usage, status visibility was limited (often only initial 👀), and reactions could remain after completion. This change improves UX by: - showing clearer progress states during execution - handling emoji compatibility limits in Telegram via fallback candidates - cleaning reactions after final reply ## What Changed ### `src/telegram/bot-message-dispatch.ts` - Added Telegram status reaction controller with phase transitions. - Switched from single emoji to ordered fallback emoji candidates per state. - Added terminal hold windows (`done` / `error`) and explicit reaction cleanup (`[]`) after completion. - Kept behavior gated to existing reaction capability path. ### `src/telegram/bot-message-dispatch.test.ts` - Added/updated tests for: - queued -> thinking -> tool -> done flow - error flow - disabled-ack path (no status reactions) - terminal cleanup call (`setMessageReaction(..., [])`) ## Behavior Notes - If a Telegram chat does not accept a preferred emoji for a state, the controller automatically tries fallback emojis. - After completion (success or failure), reaction is cleared to avoid stale status. - Existing draft streaming behavior remains unchanged. ## Testing ### Automated - `pnpm vitest run --config vitest.unit.config.ts src/telegram/bot-message-dispatch.test.ts` - Result: 17 passed / 17 total ### Manual - Verified in local Telegram runtime: - status reactions appear during processing - reactions are removed after bot reply ## Scope Focused change in Telegram dispatch only: - no schema/API changes - no unrelated subsystem refactors ## AI Assistance Disclosure - [x] Marked as AI-assisted in title/description - [x] Degree of testing noted (lightly tested) - [ ] Prompts/session logs attached - [x] Confirmed understanding of changed code paths <!-- greptile_comment --> <h3>Greptile Summary</h3> adds telegram reaction state machine with queued → thinking → tool → done/error transitions, fallback emoji support for compatibility, and automatic cleanup after completion. improves UX by showing clearer progress states during bot execution. **major changes:** - new `createTelegramStatusReactionController` manages reaction lifecycle with promise chaining - emoji fallback arrays per state (thinking, tool, coding, web, done, error) with 👀 as universal fallback - tool-specific emoji selection based on tool name tokens (web_search → 🌐, bash → 💻, etc.) - terminal states hold for 1.2s (done) or 1.8s (error) before cleanup - reaction cleanup in finally block ensures removal after completion or error **found issues:** - fallback logic has edge case where early return can skip preferred emojis when current emoji appears later in fallback list <h3>Confidence Score: 4/5</h3> - safe to merge with one logical edge case in fallback emoji handling - comprehensive test coverage (17 passing tests including state transitions, error handling, and disabled-ack path), well-structured state machine with clear separation of concerns, but contains minor fallback logic issue where early return can skip preferred emoji attempts if activeEmoji matches a later fallback candidate. implementation is sound overall with proper error handling and promise chaining. no security concerns, no breaking changes, focused scope limited to telegram dispatch only. - pay attention to `src/telegram/bot-message-dispatch.ts` lines 117-118 for the fallback logic edge case _{Last reviewed commit: 8a61d47} <!-- greptile_other_comments_section --> _{(2/5) Greptile learns from your feedback when you react with thumbs up/down!} <!-- /greptile_comment -->