Set execution options in a way more consistent with DataFusion and DataFusion Python

[robtandy]

Right now, the DFRayContext accepts options that govern its runtime behavior like

 ctx = DFRayContext(
       batch_size=batch_size,
       partitions_per_worker=partitions_per_worker,
       prefetch_buffer_size=prefetch_buffer_size,
       worker_pool_min=worker_pool_min,

While we cannot accept a SessionConfig from datafusion-python due to lack of ABI stability in rust, we should still try to adhere to the DataFusion Python API as much as possible.

So, we allow setting datafusion configuration options via DFRayContext.set.

Setting options in two places is confusion. I think it will be more clear if we set the options in the init via DFRayContext.set instead.

We can call them

• datafusion.ray.execution.batch_size
• datafusion.ray.execution.partitions_per_processor (use updated name processor instead of worker here)
• datafusion.ray.execution.prefetch_buffer_size
• datafusion.ray.execution.processor_pool_min (use updated name processor)

[jsai28]

Mind if I take a stab at this?