[BUG] reset coding env properly

[Mortimerp9]

currently, the coding env doesn't reset properly, so you can do:

env.step(CodeAction(code="def foobar():\n    return 'Hello from function'\n"))
env.reset()
env.step( CodeAction(code="print(foobar())"))

This is not correct, foobar should be removed when the env is reset.

solution: create a new PyExecutor on reset. Also reset the transform, just for good measures.

test: uv run pytest tests/envs/test_python_codeact_reset.py -v

side effect: smolagent actually requires 3.10 and it's a dependency in the pyproject, so we need to up the base python we have in the pyproject. (maybe each env should have its list of reqs in its own docker)