Add bootstrap distributions for gaps to results

[davidskalinder]

This depends on #33 (and on its dependency #31, its dependency #30, and its dependency #26); as well as on the gaps bootstrapping stuff in #32 (and on its dependencies #15 and #31, its dependencies #30, and its dependency #26 again).

I believe the only difference between this PR and #33 is that this one has #32 merged into it, which brings the bootstrapped gaps into the list of bootstrapped estimates so that the code in #33 automatically outputs them along with the others.

So after this PR, the structure of results$bootstraps looks like this:

• gaps
  • data frame with metrics (e.g., se, CI boundaries) as columns and term types (here, gap, pct_gap, etc.) as rows
• overall
  • data frame with metrics as columns and term types (e.g., endowments etc.) as rows
• varlevel
  • data frame with metrics as columns and term-type/term pairs as rows
• dists
  • gaps
  • list of term types, each containing a data frame with the column "gaps" and each bootstrap run as rows
  • overall
  • list of term types, each containing a data frame with the column "overall" and each bootstrap run as rows
  • varlevel
  • list of term types, each containing a data frame with term names as columns and each bootstrap run as rows

[davidskalinder]

Just updated this (EDIT: and in all the dependencies) so that the name of the single column in each element of both the gaps and the overall bootstrap distributions is "estimate" instead of "gaps" and "overall" (since the column "gaps" within the list element $EY_a didn't make sense).