Async memory management for OpenAIAgentWorker

[mathematisse]

Description

Following recent changes on async memory (#16127 , and other minors) this is an implementation of the async aput for the open AI agent worker.

• Added afinalize_task in the open ai agent package
• Added aput_messages in the base memory to allow for custom implementations with eventual gathers (if/when possible)
• Biggest change: Added a detection mecanism with inspect in the chat engine to ensure we call correctly the "on_stream_end_fn". That field is for now only used by open ai agent and react agent, but to ensure compatibility and non-breaking change, I did it like this. (handle detection of partial functions)

Version Bump?

• Yes, for llama-index-agent-openai

Type of Change

• New feature (non-breaking change which adds functionality)

How Has This Been Tested?

• I believe this change is already covered by existing unit tests

Suggested Checklist:

• I have performed a self-review of my own code
• My changes generate no new warnings
• New and existing unit tests pass locally with my changes
• I ran make format; make lint to appease the lint gods

[mathematisse]

I've added another await aput that was missing, my mistake this time

I've intentionally left the non async put for the memory present in extra-state, as this memory is meant to be temporary from what i understand

Merry Christmas and in advance, wish a good new year to y'all

EDIT: Mistake from my side, this aput would have applied to the extra-state memory, double mistake just corrected

If I have your approval @logan-markewich , I can also apply await logic to the extra_state memory which would standardize async use in the open ai agent, even if the BaseMemory doesn't really implement any await logic

[mathematisse]

@logan-markewich reverted vbump, merged main, ready to merge hopefully !

[mathematisse]

I've added additionnal migrations for afinalize_tasks and afinalize_response of the BaseAgentRunner, ensuring the correct functions are called