Fix lint issues and update agent.execute_task for recursion depth

- Remove unused imports and variables in test files
- Replace bare except with specific exception in structured_output_converter.py
- Fix None comparison in llm_test.py
- Update agent.execute_task to accept recursion_depth parameter

Resolves all remaining lint issues for A2A protocol implementation.

Co-Authored-By: Joe Moura <joao@crewai.com>

tests/llm_test.py

@@ -2,7 +2,6 @@ import os
 from time import sleep
 from unittest.mock import MagicMock, patch
 
-import litellm
 import pytest
 from pydantic import BaseModel
 
@@ -222,7 +221,7 @@ def test_get_custom_llm_provider_gemini():
 
 def test_get_custom_llm_provider_openai():
     llm = LLM(model="gpt-4")
-    assert llm._get_custom_llm_provider() == None
+    assert llm._get_custom_llm_provider() is None
 
 
 def test_validate_call_params_supported():