Should create_agents/AgentSet.set() deepcopy mutable per-agent values by default?

opening this per quaquel's suggestion on mesa-examples#496.

AgentSet.set() and create_agents assign per-agent sequence values as-is, no copying. mesa-examples#496 hit this directly: VirusAntibodyModel needed each of 20 virus agents to get its own dna list, and the naive way to fix the length-mismatch crash there, `dna=[dna]*20`, would have had every agent holding the same list object. VirusAgent.generate_dna mutates dna in place on reproduction (`dna[idx] = ...`), so the first virus that reproduces would've silently corrupted every other virus's starting dna too. The actual fix there is manual, wrap each element in `copy.deepcopy` before passing it in.

question is whether create_agents/set() should deepcopy mutable per-agent values (list, dict, etc) by default instead of leaving every example to know this. quaquel flagged perf as a real concern, especially for numpy arrays, so maybe this only applies to plain mutable containers rather than a blanket deepcopy, or it's opt-in via a flag. not attached to a specific implementation, just opening the discussion.

Start by reading AgentSet.set() and create_agents, then review the linked mesa-examples#496 context and VirusAgent.generate_dna's in-place mutation. Compare the behavior for mutable per-agent values with the stated performance concern around numpy arrays; done requires an agreed scope and implementation approach before coding can begin.

Written by the indexing model from the issue text.