Summary

This PR modifies the AgentSet initialization to automatically use the random number generator from the model of the first agent when no explicit random parameter is provided, eliminating unnecessary UserWarnings in typical usage.

Bug / Issue

Previously, AgentSet initialization would raise a UserWarning whenever random=None, even though in most cases the agents being added to the set already have access to a seeded random number generator through their model. This resulted in users needing to explicitly pass random=model.random in every AgentSet creation, creating unnecessary boilerplate.

Related to discussion #1981.

Implementation

The key changes made:

1. Modified AgentSet.__init__ to check if agents are provided before issuing the warning
2. Added logic to automatically extract the random number generator from the first agent's model when random=None and agents exist
3. Updated warning condition to only trigger when both the agent list is empty AND random=None
4. Expanded type hint for the random parameter to include np.random.Generator
5. Updated all test cases to remove explicit random=model.random parameters where agents are provided, demonstrating the new default behavior

The implementation maintains backward compatibility - explicitly passing a random number generator still works as before.

Testing

• All existing tests updated to use the new default behavior
• Added new test test_agentset_initialization_w_random() to verify the random number generator is correctly inherited from the model
• Verified warning is still issued when creating an empty AgentSet without specifying random
• All modified tests pass, confirming agents can access the correct seeded RNG through their model

Additional Notes

This change significantly reduces boilerplate code when working with AgentSets, as users no longer need to explicitly pass the random number generator in the common case where agents with models are being added. The warning is preserved for the edge case where an empty AgentSet is created without a specified RNG, which could lead to non-reproducible models.