Migrate some simple ddply() calls to dplyr equivalents

[MichaelChirico]

Closes #1387.

Similar to #1382, #1383, #1385, #1386 -- the end goal is moving away from {plyr}. Like #1386, this bumps {dplyr} from Suggests to Imports.

There are a variety of stylistic choices around using tidyverse code in a package (whether to use %>% / |>, whether to use the .data mask to avoid the need for globalVariables(), etc). Let me know your preferences and I can make further edits.

Also like #1386, the key driver of potential differences in these between {plyr} and {dplyr} comes down to the differences between plyr::rbind.fill() and dplyr::bind_rows(). Since the structure across groups in summarize() expressions is typically consistent, I don't have much concern about that here.

The second major difference is that plyr::ddply() always sorts the output by the grouping key, whereas dplyr::summarize() sorts according to the input row order (this in turn matches the {data.table} behavior). I tried to have a look at nearby code to see if the output order matters -- if it's not clear we can ignore the row ordering, I default to assuming it's important, thus we might be able to drop more arrange() calls.

There are still a large number of ddply() calls -- this reduces the number from roughly 74 to roughly 48. The remaining ones either do more complicated things around column selection, or use a much more complicated .fun which requires more careful examination. More PR(s) to follow.