Create mask array which marks certain values to be excluded from surrogate training

[Lance-Drane]

High level need

It's possible that instruments/users/etc. may generate bad values and not want them to be included in training, but they would like to archive the values. It would also serve as a way to represent a "hole" in the dataset.

This also makes INTERSECT orchestration easier, both for campaign authors and the

Low level implementation

• Represent the mask in the database, generally encapsulate this from the user
• allow users to initialize_workflow with masked value(s) (provide optional array of indexes with masked values)
• allow users to update workflow with bad value(s) (simple boolean flags)
• store both an all_values dataset_x + dataset_y and a good_values dataset_x + dataset_y in the database for efficiency