Functionality for Generalized Spatial Temporal Areal Aggregated Predictors

[apeterson91]

Dependent on benvo data structure that accommodates sf objects and associated characteristics,
create a new keyword e.g. saap() that accommodates subject-associated areal aggregated exposures.

As a first step could fix the predictor as a tensor product of space and area and then find a good method for generalizing to arbitrary features. For example BMI ~ sex + year + saap(Parks) will automatically fit an aggregated smooth function of subject distances and the park's corresponding distance. Could then eventually make something like BMI ~ sex + year + saap(Parks,Distance,TreeCover,Area) that would allow for a three dimension smooth as a function of subject-park distance, park tree cover and park area.