Make `epi_df` compatible with `rowwise`

[brookslogan]

Rowwise operations on epi_dfs lose epi_df-ness.

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(epiprocess)
#> 
#> Attaching package: 'epiprocess'
#> The following object is masked from 'package:stats':
#> 
#>     filter
toy_epi_df = new_epi_df(tibble(geo_value=1, time_value=1:5, cases=1:5))
toy_epi_df %>% rowwise() %>% mutate(geo_value2 = geo_value)
#> # A tibble: 5 × 4
#> # Rowwise: 
#>   geo_value time_value cases geo_value2
#>       <dbl>      <int> <int>      <dbl>
#> 1         1          1     1          1
#> 2         1          2     2          1
#> 3         1          3     3          1
#> 4         1          4     4          1
#> 5         1          5     5          1
toy_epi_df %>% rowwise() %>% group_split()
#> <list_of<
#>   tbl_df<
#>     geo_value : double
#>     time_value: integer
#>     cases     : integer
#>   >
#> >[5]>
#> [[1]]
#> # A tibble: 1 × 3
#>   geo_value time_value cases
#>       <dbl>      <int> <int>
#> 1         1          1     1
#> 
#> [[2]]
#> # A tibble: 1 × 3
#>   geo_value time_value cases
#>       <dbl>      <int> <int>
#> 1         1          2     2
#> 
#> [[3]]
#> # A tibble: 1 × 3
#>   geo_value time_value cases
#>       <dbl>      <int> <int>
#> 1         1          3     3
#> 
#> [[4]]
#> # A tibble: 1 × 3
#>   geo_value time_value cases
#>       <dbl>      <int> <int>
#> 1         1          4     4
#> 
#> [[5]]
#> # A tibble: 1 × 3
#>   geo_value time_value cases
#>       <dbl>      <int> <int>
#> 1         1          5     5

^{Created on 2023-03-15 with reprex v2.0.2}

We may want the first case to work if we have more complicated data in a list column to manipulate with non-vectorized functions. We may want the latter to work, e.g., to simplify internals of #275 when faced with 0 vs. non-0 row epi_dfs, though this probably also means implementing some stuff in pillars and/or vctrs.

[brookslogan]

toy_epi_df[integer(0),] %>% group_by(geo_value) %>% group_split() outputs

<list_of<
  epi_df<
    geo_value : double
    time_value: integer
    cases     : integer
  >
>[0]>

so maybe there is not any pillars/vctrs work to do.