Fix toolInUse initialization to correctly detect tool usage

[Nepomuk5665]

Summary

This PR fixes a bug in bedrock_llm_agent.py where the toolInUse variable was incorrectly initialized to True before the loop that checks for tool usage, making it always True regardless of whether a tool was actually used.

Bug Description

In both handle_single_response and handle_streaming_response methods:

# Before fix (incorrect)
toolInUse=True  # Bug: always True
for item in response_content:
    if isinstance(item, dict) and "toolUse" in item:
        content.append(item)
        toolInUse = True  # Redundant, already True

This caused reasoningContent (thinking content) to always be inserted at index 0 of the content array, even when no tool was being used. According to the code comment, reasoning content should only be inserted at index 0 when a tool is used, and appended otherwise.

Fix

• Changed toolInUse = True to toolInUse = False to correctly initialize the variable
• Added a check for thinking_content/accumulated_thinking before appending to avoid appending None values when there is no thinking content
• Updated comment to correctly describe the loop's purpose (saving "toolUse items" not "text items")

Impact

When extended thinking is enabled and no tool is used, the reasoningContent will now be correctly appended to the content array instead of being inserted at the beginning.