automatminer_dev workflow improvements

[ardunn]

E.g., a method of MatPipe.

Takes in execution options (e.g., featurization and tpot in separate fireworks vs. everything in one firework) and returns a fireworks workflow to execute it. Should also take in the path to a data file on the executing computer resource, to actually run the workflow on.

Edit:

todo

• [ ] add a workflow for predicting properties to benchdev

[ardunn]

could be simple (i.e., not as complex as what is on hmprivate)

def get_benchmarking_wf(matpipe_config, data_file, target, kfold, save_dir):
    returns wf

Then just saves all dfs, pipes, logs, digests, and text summaries in hmprivate style in the save_dir

[ardunn]

merging this issue with #227

[ardunn]

Todo:

• separate the workflow into one firework for each of the 4 operations
• make and test a workflow for single fits