Many agent applications often benefit from being able to retrieve data from memory (rag).
This PR is meant to enable ..

• Defining memory as a component that can be declaratively specified and attached to an agent, similar to how a model or tool can be attached to an agent
• Implement several base Memory classes e.g, Memory as list,

Design Considerations.

Memory is by nature should be persistent across runs - using some type of db.

• Declarative spec will keep a pointer to the db entry
• Have flexible defaults - e.g., create a memory entry if the specified one is unavailable etc. not fail etc

Should build on #4039 #4100