New BubbleML Dataset Tasks

[arthurfeeney]

Some work needs to be done before the new dataset can really be generated.

• x-locations of nucleation sites.
  • ideally, Flash-X will just spit out the x-locations, so we can always match the respective simulation.
  • Can be input to the model using the sigmoid of an SDF from the x-locations.
• ability to change random seed for simulations. I want possible duplicate temperatures, but with different seeding locations. (this is probably already possible).
• meta-script to generate the flash-x config files.
  • not strictly needed, but annoying to manually edit a ton of files :-)
• Update pre-processing script. Since the format of the data has changed some, can't directly use datagen/bubbleml/construct_train_test.py
• run a ton of simulations :-)