Add support for the sample model_output_type in hubEnsembles::linear_pool()

[nickreich]

Add support for the sample model_output_type in hubEnsembles so that hub administrators and/or individual modelers can run linear pool ensembles on model_output in sample format.

Acceptance criteria:

• sample output_type can be passed to linear_pool() with valid ensemble output returned
• unit tests in place to check accuracy of calculations
• documentation of function, including arguments and explanations for any analysis choices that need to be specified by the user
• warning is thrown if the number of samples available for all prediction units and from all models are not equal

[elray1]

We should discuss and combine this issue with #27

[elray1]

r.e. "warning is thrown if the number of samples available for all prediction units and from all models are not equal" -- is that something that is a part of an ensembling operation? feels like a validation that should happen somewhere else?

[nickreich]

I think if a user passes model output that has varying numbers of samples per prediction unit (maybe excluding 0s?) there should be a warning thrown. It feels like the kind of thing that could be overlooked but could have weird consequences depending on the implementation.