Think through solver interface for better serialization with ray

[aadharna]

We can parallelize better at this line:

watts/watts/transfer/score_strategy.py

Line 26 in 8c1ce43

refs.append(s.evaluate.remote(env_config=self.env_config,

By using the Solver system we are forcing a serial call for each actor, but are parallelizing by running each actor independently.

Think through better ray usage?

[aadharna]

Ray links to two instances of additional concurrency when using Actors. So, that's probably the best place to start looking -- https://docs.ray.io/en/latest/actors.html#concurrency-within-an-actor