Enable custom cluster managers

[MilesCranmer]

This let's PySR launch jobs over a cluster. If you are on a slurm cluster (on the head node of a job), you simply pass cluster_manager="slurm" to PySRRegressor, and it will set up a bunch of workers over the slurm job. Likewise, the other cluster managers listed on https://github.com/JuliaParallel/ClusterManagers.jl can be used.

The one unfortunate thing is I don't think this functionality can be unit-tested; it will just be manual tests on clusters.

cc @kazewong!

[MilesCranmer]

This model:

model = PySRRegressor(
     unary_operators=["cos"],
     procs=40 * 4,
     populations=40 * 4 * 2,
     population_size=1000,
     ncyclesperiteration=3000,
     cluster_manager="slurm",  # < Only change required
)

will launch 160 workers over a slurm job (4 nodes in this case), when ran from the head node.

Here's what the usage looks like:

Keep in mind this was launched from within IPython on the head node!