Add row group id as an argument to tab_row_group()

[rich-iannone]

This PR seeks to do what #665 did recently, which is to require and use IDs for different row groups. The function signature for tab_row_group() will be similar to that of the analogous tab_spanner():

tab_row_group(
  data,
  label = NULL,
  rows = NULL,
  id = label,
  others = NULL
)

This is not the only way to specify row groups in a gt table. Here is a breakdown of the three different ways to include them:

1. with grouped data given to gt(); group labels are composed by the group column names being pasted together, each separated by _
2. using the groupname_col arg in gt(); we can only choose one column so flexibility in grouping is purposefully limited compared to method (1)
3. by use of the tab_row_group function. This is used past the initial gt() step.

Aside from introducing IDs here, this PR aims to fix a number of known bugs along the way.

Fixes: #671
Fixes: #675
Fixes: #717

[jcheng5]

Just to note a limitation on the column splitting logic, if delim contains "\\E" then the behavior will be wrong.

split_at_first_delim <- function(delim, str) {
  delim <- gsub("\\E", "\\\\E", delim, fixed = TRUE)
  m <- regexec(perl = TRUE,
    paste0(
      "^(.*?)\\Q", delim, "\\E(.*)$"
    ),
    str
  )
  regmatches(str, m)
}

split_at_first_delim("\\E", "foEo\\E.bar")  # expected: foEo|.bar
split_at_first_delim("\\", "foo\\bar")      # expected: foo|bar
split_at_first_delim("\\\\E", "foo\\\\Ebar")      # expected: foo|bar

We could fix this, or, simply detect \\E in the delim and throw. I don't think anyone will do this in good faith, but this will prevent any weird attack vectors(?).