Make print.epi_df header mirror style of print.tsibble and print.tbl_df

[brookslogan]

Currently, print.epi_df's "header" looks like

An `epi_df` object, with metadata:
* geo_type  = state
* time_type = day
* as_of     = 2020-06-03

# A tibble: 9 × 4

(from `ex1 from #103) or

An `epi_df` object, with metadata:
* geo_type  = state
* time_type = day
* as_of     = 2022-05-23 13:17:07

# A tibble: 4,026 × 6

(from jhu_csse_daily_subset).

In editors that color the # line, it makes it easy to overlook the epi_df part of this and think it's just a tibble.

Edit this header to be something more like

# An `epi_df` object: <nrow> x <ncol>, with metadata:
# * geo_type  = state
# * time_type = day
# * as_of     = 2020-06-03

which improves the user experience if they have that #-line-coloring editor behavior, and to match the style of tibble and tsibble printing.

[brookslogan]

Additionally, we should print additional_metadata, or at least other_keys. Moved to #186

[brookslogan]

Reopening this, as the current main only has part of the changes; current main print looks something like:

An `epi_df` object, 4,026 x 6 with metadata:
* geo_type  = state
* time_type = day
* as_of     = 2022-05-23 13:17:07

# A tibble: 4,026 × 6
[rest of tibble output]

which would look better as something like

# An `epi_df` object, 4,026 x 6 with metadata:
# * geo_type  = state
# * time_type = day
# * as_of     = 2022-05-23 13:17:07
[rest of tibble output]

or

# An `epi_df` object, 4,026 x 6 with metadata:
# * geo_type  = state
# * time_type = day
# * as_of     = 2022-05-23 13:17:07

[rest of tibble output]

especially when using an editor that colors the "comment" lines (approximated above by using perl as the "language" of the code blocks to trigger coloring of the comment lines and not a bunch of coloring of the other lines). Currently, one's eyes could be drawn to these "comment" lines and users may read too quickly and think that an epi_df they are printing is just a tibble.

Still just a P3, so it shouldn't take precedence over any the epipredict work (P0).

[brookslogan]

Additionally, it looks like print on current main may still not be printing the other_keys metadata. Moved to #186

[brookslogan]

Based on discussion in #183, I'm opening another issue about the other_keys printing, and removing it from this one.

[brookslogan]

Think we might need to provide a tbl_format_header and/or tbl_sum method for this (generics are in the pillar package). [And check for interop with grouping. Problematic as grouped_df expects to be in front of the "base"-ish class in the class vector so it can tack stuff onto NextMethod(), but we currently put the "base"-ish class epi_df in front of grouped_df. That swap might be fine now, but might potentially cause issues for some other dplyr verbs, especially if we pursue #295; we'll need to check.]