Add in customization options for local cluster resources

For instances where a Dask cluster is necessary, the user should be able to specify the amount of cores, RAM, etc. to use for the cluster. Right now, it defaults to creating a cluster with 1 process and 1 thread but uses all available memory which is not ideal. Also need to be able to change the compute fraction reserved for GPU compute when estimating whether a dataset can fit in VRAM fully without needing to use Dask for partitioning.

assigned to @mdefende

marked this issue as related to #32 (closed)

mentioned in merge request !56 (merged)

closed with merge request !56 (merged)