Migrate documentation to roxygen2

[MichaelChirico]

Noting some inconsistencies/gaps for which I couldn't find a workaround, and intentional splits:

• NA_integer64_ gets some extra metadata Disable behavior of @export to _always_ treat a non-function as a "data" documentation roxygen2#1666
• Using @describeIn puts those annotations in a \section{Functions} instead of the \description{}
• Markdown **text** translates to \strong{text}, not \bold{text}, AFAICT there's not too much meaningful difference
• is.integer64 was being documented as an S3 method (it's not registered as such in NAMESPACE); probably this is a relic of 1d34b37:

bit64/man/bit64-package.Rd

Line 55 in a014e88

\method{is}{integer64}(x)

bit64/NAMESPACE

Lines 371 to 373 in a014e88

	#no registration because of signature mismatch
	#S3method("is", integer64)
	#S3method("as", integer64)

• {roxygen2} did a great job of catching a case where %in% was not properly escaped and we wound up with unintentionally-commented exposition (note the awkward '):

bit64/man/bit64-package.Rd

Line 36 in a014e88

'match', '%in%', 'duplicated', 'unique', 'table', 'sort', 'order', 'rank',

• Drop \details{...} section from bit64-package which does not add much and is already way out of date:

bit64/man/bit64-package.Rd

Lines 79 to 89 in d783d63

	\details{
	\tabular{ll}{
	Package: \tab bit64\cr
	Type: \tab Package\cr
	Version: \tab 0.5.0\cr
	Date: \tab 2011-12-12\cr
	License: \tab GPL-2\cr
	LazyLoad: \tab yes\cr
	Encoding: \tab latin1\cr
	}
	}

Some code to compare NAMESPACE across branches (normalizing away trivial differences like sort order):

parse_namespace_on_branch = function(branch) {
  system(paste('git checkout', branch))
  wd = getwd()
  parseNamespaceFile(basename(wd), dirname(wd))
}
old = parse_namespace_on_branch("master")
new = parse_namespace_on_branch("roxygen")

normalize_imports = function(imports) {
  df = do.call(rbind, lapply(imports, \(e) data.frame(pkg = e[[1]], calls = e[[2]])))
  df = df[order(df$pkg, df$calls), ]
  row.names(df) = NULL
  df
}

normalize_s3_methods = function(s3) {
  df = data.frame(s3)
  names(df)[1:2] = c("generic", "class")
  df = df[order(df$generic, df$class), ]
  row.names(df) = NULL
  df
}

for (nm in names(new)) {
  normalizer = switch(nm,
    imports = normalize_imports,
    exports = sort,
    S3methods = normalize_s3_methods,
    identity)
  stopifnot(all.equal(normalizer(old[[nm]]), normalizer(new[[nm]])))
}