compute list of predictions given a dataset and list of selections

[kirangadhave]

If we decide on the shape of the prediction object, it would be helpful to generate Python type-hints to go along with Typescript types.

The prediction object should include the following:

name - display name
algorithm - the name of the exact algorithm
params - params that are required to rerun the algorithm
dimensions - list of dimensions considered during computation
output - list of points selected by the algorithm
rank - similarity score compared to the user selection
metadata - extra data that would be algorithm-specific. This is used to render the output on the screen. e.g., for linear regression, it includes the slope and intercept.

The dataset can be a dataframe for now, but we should ideally accept json, file paths, etc. and convert it to dataframe

The list of selections would be indexes of the dataset OR values of unique id field. I will have to think about this, but indexes should be fine for now