Create a frequency cache

[jqnatividad]

Similar to the stats --stats-jsonl option, create a frequency --frequency-jsonl option.

The frequency cache will have the COMPLETE frequency table for each column, not just the top N (default: 10) values, as is the default behavior of frequency.

This frequency metadata will allow us to support more "smarter" use cases that require more info than the stats cache provides.

[jqnatividad]

IMPLEMENTATION NOTE:
The frequency cache also depends on the stats cache existing before its computed so it will only have an ALL_UNIQUE entry for ID columns.

Also, for columns with cardinality > 1000 or 90% (both configurable) of the rowcount, whichever is smaller, it will have HIGH_CARDINALITY instead.

In this way, the frequency cache should still be relatively "tiny" metadata, even for very large CSV files.