Fix checks: "R code for possible problems", "Rd files"/"metadata"/"cross-references"/"code/documentation mismatches"; kml_to_csv bug/tests

[mhpob]

This PR is an attempt to start clearing up some of the "ignorable" output of R CMD check. Commits are small to aid in cherry-picking.

• Declares all of the variables used in dplyr, data.table, or base (subset) NSE (Closes dplyr programming throws R CMD check warnings #190)
  • This was done using the data.table suggestion rather than the tidyverse rlang ".data" way as it seems a universal fix for all programming styles and didn't add a dependency.
  • Necessary to pass R CMD check: "checking R code for possible problems"
• Explicitly calls dplyr::across, stats::setNames, utils::head, and utils:tails
  • Necessary to pass R CMD check: "checking R code for possible problems"
• Fixes bug in kml_to_csv (Closes kml_to_csv doesn't check file extension as expected #227) and adds a few unit tests for kml_to_csv
  • Former is necessary to pass R CMD check: "checking R code for possible problems"
• Adjust target to "_PACKAGE" (introduced in d5d5759) and remove LATeX escapes in package-glatos.R
  • Necessary to pass R CMD check: "checking Rd files"
• Fixes misc. mis-specified itemized list in documentation
  • Necessary to pass R CMD check: "checking Rd files"
• Puts backticks around anything with [...] that isn't supposed to be a link
  • Passes "checking Rd cross-references"
• Switch @usage roxygen tag to @example in testdata.R and extdata.R to get around the need to document system.file
  • Necessary to pass R CMD check: "checking for code/documentation mismatches"
• Slight adjustment of convert_otn_erddap_to_att test from expect_identical to expect_equal to avoid failure related to floating point issues on macOS