Duplicate CrewAI Workflow Execution Causes Excessive LLM/API Calls

**Describe the bug**

A double execution issue exists in the Slack background worker task `generate_ai_reply_if_unanswered`, causing the full CrewAI multi-agent workflow to execute twice for a single user query.

The issue occurs because after generating an AI response using `process_ai_query(query=message.text)`, the generated response (`ai_response_text`) is passed as the first argument to `get_blocks()`:

```python
client.chat_postMessage(
channel=channel_id,
blocks=get_blocks(ai_response_text, channel_id=channel_id),
text=ai_response_text,
thread_ts=message.slack_message_id,
)
```

However, `get_blocks()` treats its first argument as a new user query and internally calls `process_ai_query()` again:

```python
ai_response = process_ai_query(
query.strip(),
images=images,
channel_id=channel_id,
is_app_mention=is_app_mention
)
```

This causes the complete CrewAI pipeline (Query Analyzer → Router Agent → Expert/Collaborative Agents → Synthesizer) to run a second time using the bot's generated response as input.

Impact:

* Doubles LLM/API usage for every background-processed query
* Increases latency and operational cost
* May produce incorrect formatting or clarification-style responses because the generated markdown response is interpreted as a new user question

## Expected behavior

The CrewAI workflow should execute only once per user query.

`get_blocks()` should format the already-generated response without re-processing it through `process_ai_query()`.

---

## Are you going to work on fixing this?

* [x] Yes
* [ ] No

## Additional context

This issue becomes significantly more expensive with collaborative flows and query decomposition enabled because a single query may already trigger multiple agent/tool executions. The unintended second execution compounds token usage, latency, and overall API cost across the entire CrewAI pipeline.