[YSQL] Improve PopulateParallelSelectOps() to allow partition pruning

[andrei-mart]

Jira Link: DB-3331

Description

The PgDocReadOp::PopulateParallelSelectOps() method is designed to create multiple copies of the read request and bind them to different partitions, so they can be executed in parallel. However, it creates requests to all partitions, while certain conditions allow to exclude certain partitions. Currently we prefer pruning, so if request has conditions allowing to filter out (prune) partitions, we do not do parallelism, so if request targets multiple partitions, they are scanned sequentially.
We should be able to do both, pruning and parallel requests to the remaining partitions.