join (via sql): add --all option to join on all key values, not just those in the first input

[liquidaty]

For example, given 2 inputs:

a.csv:

id,..
1,...
2,...

and b.csv:

id,..
2,...
3,...

then sql --join-indexes 1 a.csv b.csv returns two data rows with ids 1 and 2, but sql --join-indexes 1 --all a.csv b.csv would return three data rows with ids 1, 2, and 3