Handle exceptions in parallel agent runs to prevent early termination

[mshsheikh]

Problem:
When running agents in parallel using asyncio.gather, any unhandled exception in one task (e.g., network error, agent crash) would immediately propagate and terminate the entire workflow, discarding all partial results.

Example of Failure:

res_1, res_2, res_3 = await asyncio.gather(
    Runner.run(spanish_agent, msg),  # Fails with Exception
    Runner.run(spanish_agent, msg),
    Runner.run(spanish_agent, msg),
)
# All results are lost due to a single failure.

Changes:

• Added return_exceptions=True to asyncio.gather() to handle failures gracefully:

  res_1, res_2, res_3 = await asyncio.gather(
      Runner.run(spanish_agent, msg),
      Runner.run(spanish_agent, msg),
      Runner.run(spanish_agent, msg),
      return_exceptions=True,
  )

• Exceptions are now captured as Exception objects in the results instead of being raised immediately.

Benefits:

• Ensures robustness: Partial results are preserved even if one or more tasks fail.
• Allows explicit error handling for failed tasks (e.g., logging, fallback logic).
• Prevents unnecessary re-runs of successful tasks when one task fails.

Example Output Handling:

outputs = [
    ItemHelpers.text_message_outputs(res_1.new_items) if not isinstance(res_1, Exception) else None,
    # ...similar for res_2, res_3
]