Implement percent_rank and cume_dist properly to ibis.omniscidb

[xmnlab]

Ibis percent_rank doesn't implement the correct SQL percent_rank algorithm, instead it implements the algorithm for cume_dist, the same behavior of pandas (pandas-dev/pandas#28975).

Currently, ibis core has already percent_rank, so probably we can implement the percent_rank in the SQL way for now here.

for the cume_dist, maybe we can create the operator class here first, and open a PR (in parallel) to ibis core.

The work made at PR ibis-project/ibis#2224 can be used as reference/base.