It's interesting that there doesn't seem to be a command-line tool out there that lets you do this with LevelDB. LevelDB is about an order of magnitude faster at importing a lot of data than SQLite or Postgres is (although I don't know if it's faster than the Postgres CSV foreign data wrapper for the first query, it sure will be for the second) and it can handle files a lot bigger than your RAM.
It's interesting to note that around 1990 The Mark Williams Company wrote a fairly full-fledged database system using, basically, CSV files. It was called "/rdb". You wrote your queries as shell scripts, but using a set of /rdb utilities that handled CSV files with labeled columns, and they had a screen-based data-entry UI based on vi. I wouldn't want to use it instead of a database --- you had to write your query plan in the shell script, rather than a high-level SQL-like query, because the authors didn't really understand SQL --- but it's interesting that the approach is still useful in 2019.
It's interesting to note that around 1990 The Mark Williams Company wrote a fairly full-fledged database system using, basically, CSV files. It was called "/rdb". You wrote your queries as shell scripts, but using a set of /rdb utilities that handled CSV files with labeled columns, and they had a screen-based data-entry UI based on vi. I wouldn't want to use it instead of a database --- you had to write your query plan in the shell script, rather than a high-level SQL-like query, because the authors didn't really understand SQL --- but it's interesting that the approach is still useful in 2019.