improved handling of CSV comments

[missinglink]

as mentioned in https://github.com/pelias/geonames/issues/404#issuecomment-1055506176 there seems to be a weird bug with how the geonames metatadata files are encoding comments, (I think!)

using this custom comment handler stream we're able to work around the issue, although I'm still not clear why the comment option from https://csv.js.org/parse/options/ (and sed '/^#/d') doesn't do the same thing 🤷

I've also taken the opportunity to do some simple housekeeping tasks:

• upgrade csv-parse module
• enabled bom option for csv-parse as we have done in other modules
• remove the mkdirp module introduced in https://github.com/pelias/geonames/pull/185, since that time we've consolidated on Docker and Windows has made progress in its terminal utilities, I hope it's no longer required.
• change the engines definition in package.json from >=l2.0.0 to >=12.0.0, @orangejulius is this just a typo?

The 'actual work' here is:

• setting comment: '' in the csv options to disable stripping comments within that lib
• adding split2 and the new through streams to handle this ourselves.

resolves https://github.com/pelias/geonames/issues/404

[missinglink]

agh woops, so the xsv failure was my fault since I wan't explicitly telling it the file was TSV instead of CSV:

curl -s http://download.geonames.org/export/dump/countryInfo.txt | sed '/^#/d' | xsv cat -d '\t' rows

I suspect there's just a weird bug in csv-parse

[missinglink]

opened an issue upstream https://github.com/adaltas/node-csv/issues/325 hopefully we can remove these commits if a solution can be found within that lib natively.