adding tracing support for bing custom search tool

[M-Hietala]

adding tracing support for bing custom search tool

[Copilot]

Pull request overview

This PR adds tracing support for Bing Custom Search tool in the Azure AI Projects SDK. The changes enable proper telemetry instrumentation to track both Bing Custom Search tool calls and their outputs, distinguishing them from similar remote function call patterns.

Key Changes

• Added a new _emit_tool_output_event helper method to emit tool output events with the appropriate role and event type
• Updated dictionary keys for output_items to use (id, type) tuples instead of just id to prevent call and output items with the same ID from overwriting each other
• Enhanced the generic tool call/output handler to support Pydantic v2 model_dump() method in addition to existing as_dict() support
• Split handling of remote_function_call (the call) and remote_function_call_output (the output) into separate code paths with appropriate event types and field extraction logic

[Copilot]

[nitpick] These debug comments appear to be leftover from development and don't add value to the code. They should be removed to keep the codebase clean.

Suggested change

	# Debug sync wrapper initialization
	# Return wrapped stream

[Copilot]

The conditional expression "id" if id_field == "id" else "id" always evaluates to "id" regardless of the condition. This means when id_field is "call_id", the code still assigns to tool_output["id"] instead of mapping call_id to id. This appears to be a copy-paste error.

This should be: tool_output["id"] = getattr(output_item, id_field) to correctly handle both field names.

Suggested change

	tool_output["id" if id_field == "id" else "id"] = getattr(output_item, id_field)
	tool_output["id"] = getattr(output_item, id_field)