object class to represent a collection of forecasts

[elray1]

Which of the options in R for dataframe-like objects do we want to build on for a collection of forecasts that have been loaded into an R working session: data.frame, tibble, or data.table?

Here are some points that have been raised in previous discussion:

• data.frames are the only of these options that are native to R.
• tibbles and the tidyverse are a de facto standard that is in common use by a large swath of the R community, and they allow for features that are not supported in data frames such as nesting
• data.tables are substantially faster than tibbles
• We could support multiple options... but ultimately, even if we go this route we are probably going to want to pick one that most of our tooling works with, right?

[annakrystalli]

This issue is related to #33 and #66. Given arrow by default returns tibbles (so is an indirect dependency anyways) I vote for tibbles, both for the nesting and the much nicer print methods also.