fix(tui): provide fallback response when subagent produces no text output

[echobt]

Problem

When a subagent (Task tool) executes tool calls but the LLM produces no textual response (only tool calls), the parent agent previously received an empty response with just metadata. This caused scenarios where:

1. The research subagent executed tools (TodoWrite, LS, Read)
2. A Batch call might have failed
3. No textual report was included in the response - the agent sent nothing back

The user observed this as: "L'agent n'a rien envoyé" (The agent sent nothing)

Root Cause

In spawn_subagent(), the final_content variable captures the LLM's text output from each iteration. However:

• If the LLM only produces tool calls without any text explanation (Delta events)
• The iteration_content remains empty
• final_content is only updated when !iteration_content.is_empty()
• Result: The output contains only metadata, not a meaningful response

Solution

Added a fallback mechanism that provides meaningful feedback when final_content is empty:

1. If no tools AND no content: Returns a message indicating potential issues with the task/model
2. If tools executed but no content: Returns a summary with:
   • Number of tool calls
   • Success/failure counts
   • The original task description

This ensures the parent agent always receives actionable feedback from subagents.

Changes

• Modified spawn_subagent() in cortex-tui/src/runner/event_loop.rs
• Added effective_content variable with fallback logic
• The output now uses effective_content instead of final_content

Testing

• cargo check -p cortex-tui passes
• cargo test -p cortex-tui event_loop passes (9 passed, 7 ignored)
• Unrelated test failures (file icons) are pre-existing