Enhance ReActAgent's behavior with new messages during invocation

[coderabbitai]

Current Behavior

When new messages arrive during ReActAgent's invocation, they are currently queued for later processing.

Desired Enhancement

Consider implementing alternative strategies for handling incoming messages during agent invocation:

1. Interrupt current invocation: Drop the current invoke to respond to new data as quickly as possible. This would prioritize responsiveness to new information.

2. Prioritize current processing: Option to drop new messages when the agent is busy, prioritizing the completion of the current task.

Benefits

• More flexible message handling would improve the agent's responsiveness in time-sensitive scenarios
• Allow for different operation modes depending on the use case (e.g., interactive vs. batch processing)
• Provide more control over the agent's attention management

Implementation Considerations

• Add configuration options to specify the preferred behavior
• Consider how to gracefully interrupt ongoing processing
• Implement appropriate notification mechanisms when messages are dropped

Ref: PR #450

[boczekbartek]

@maciejmajek I think this issue might be resolved #538. See new_message_behavior argument of LangChainAgent:

rai/src/rai_core/rai/agents/langchain/agent.py

Line 111 in 20b14b4

new_message_behavior: newMessageBehaviorType = "interrupt_keep_last",

[maciejmajek]

That's correct, closing