Better summary composition scheme for multiple API function calls

[nischalshrestha]

Description

Often times, R code (especially tidyverse) will have nested functions being called. For example, here:

mtcars %>%
  names() %>%
  map(~count(mtcars, .data[[.x]]))

the map function call also has a count function call as part of the lambda function to apply to each of the names(). This means that we can have potentially a 'nesting' of summaries where the inner function call (count()) summary needs to be somehow incorporated with the outer (map()).

Currently, the tidylog side and the way we've cached summaries to access them (instead of display()'ing) prevents us from stitching the two summaries together. Instead, the count summary overwrites the map since map also does not have summary support yet (issue #97).

We need a more principled approach to gather summaries such that we can capture--to start--the API function call summaries and unify them into one summary for the most outermost function.