`stats`: add dataset level stats

[jqnatividad]

Currently, stats only computes column-level stats.

Also add dataset-level stats, with the "_qsv_" prefix, like:

• rowcount (_qsv_rowcount)
• column count (_qsv_columncount)
• filesize (_qsv_filesize)

• file hash (_qsv_hash) using xxHash algorithm (use twox-hash crate)

  The value for each dataset stat will be stored in a column named _qsv_value

[jqnatividad]

instead of computing the file hash which may take a long time for large files, just compute the hash of all the stats, including the rowcount, column count and the filesize.

This pretty much guarantees the hash will be unique for the file in its current state, without having to scan the entire file, serving as a "fingerprint hash."

[jqnatividad]

2297 is still WIP, but changed the dataset-level stats to:

• qsv__rowcount
• qsv__columncount
• qsv__filesize_bytes
• qsv__fingerprint_hash with their values stored in the last column of stats as qsv__value.

Removed the leading underscore because it was tripping up select in CI as underscore is a select sentinel value for last column. Made the prefix qsv__ with two trailing underscores.

[jqnatividad]

This is not truly done until corresponding CI tests succeed. stats has hundreds of tests, so this will take a bit of effort.