fix(langgraph-checkpoint-aws): Add max_results attribute to AgentCoreMemorySaver

[michaelnchin]

Addresses #719.

This PR updates the AgentCore memory checkpointer integration to:

• Add an max_results attribute to AgentCoreMemorySaver for limiting the number of events processed.
  • Currently, the limit is hardcoded to 100 events - this will remain as the default for the new attribute. None can also be specified to set no upper limit.
• Fix the limit parameter of AgentCoreMemorySaver.list(). In BaseCheckpointSaver.list(),
  • In the base method, limit configures the max# of checkpoints to return, but we also use this in the get_events() call. Updated this call to use the new class attribute limit instead.
• In AgentCoreEventClient,get_events(), exit from the response processing loop early if the number of events found exceeds max_results. If this occurs, a warning will be returned to the user.

[michaelnchin]

Hi @ltoniazzi, thanks for the suggestions in #719 - feel free to try out these changes and let me know if you have any additional feedback.

[michaelnchin]

@jgordley @bergjaak - ready for review

[ltoniazzi]

I wonder if it should default to None instead of 100, as I think saver always tries to list events for a thread_id, so I do not see a reason why one would not want to fetch the full checkpoint history by default (also limit in other savers defaults to None as well).

Also, I'm a bit surprised that this change does not break any unit test (I can try add some next week)

[michaelnchin]

@ltoniazzi actually, looking at the ListEvents API reference, maxResults has a hard upper limit of 100 events (defaults to 20 if we don't pass it). I think this is the reason for the original hardcoded limit. If the docs are correct, I think we need to do away with the no-limit None option.