SQL ungrouped summarize should always return 1 row

[machow]

Note that currently, if you do an ungrouped summarize, no GROUP BY is created.
However, this creates the issue where assigning a literal in a summarize, will be the length of all rows in table.

Reduces to one row

summarize(tbl, y = n(_)) >> show_query()
SELECT count(*) AS y 
FROM siuba_019

Does not

summarize(tbl, y = 1) >> show_query()

[machow]

I tried adding a trivial grouping operation, in the case where there are no groups. Worked on postgres butfailed at least on sqlite. Checking dbplyr's behavior, it looks like this is an outstanding issue there, too. Going to remove from sprint.

Example query:

SELECT 1 as y
FROM users GROUP BY TRUE