Move reward function out of simulator

[kasanari]

To avoid accidental leakages of statefulness.

Should also make it easier to define custom reward functions.

[mrkickling]

Made this a draft before the pipeline checks suceeed.

• Run ruff format and ruff check to fix ruff issues
• fix mypy issues

malsim/mal_simulator/rewards.py:6: error: Incompatible import of "Generator" (imported name has type "type[numpy.random._generator.Generator]", local name has type "type[typing.Generator[_YieldT_co, _SendT_contra, _ReturnT_co]]")  [assignment]
malsim/mal_simulator/rewards.py:50: error: Missing type parameters for generic type "Generator"  [type-arg]
malsim/mal_simulator/rewards.py:78: error: Argument 1 to "sample_value" of "TTCDist" has incompatible type "typing.Generator[Any, None, None]"; expected "numpy.random._generator.Generator | None"  [arg-type]
malsim/mal_simulator/simulator.py:124: error: Argument "rng" to "attacker_step_reward" has incompatible type "numpy.random._generator.Generator"; expected "typing.Generator[Any, None, None]"  [arg-type]

• pytests failing (maybe something changed in how the reward is calculated?)

[kasanari]

I ran ruff, there is a commit for it. (Update I ran the wrong ruff)

The tests are failing on the master as well. (I get them to pass now, so I may have ran something wrong. I will check the failing tests.)

[mrkickling]

I'll take a look.

[mrkickling]

Tests now passing on main using latest mal-toolbox. Try rebasing and you should see which tests started failing in this PR.

[kasanari]

This PR is also outdated since I realized the reward functions were not behaving as I expected.