Implement random/targeted bucketing for cte materialization

[jaystarshot]

Currently the first projection column is taken as the bucketing column for cte materialization. However it makes more sense to implement random bucketing instead since that will remove skew from readers and writers and improve performance.
We could also look for downstream exchanges and add bucketing on that exchange column and potentially remove the exchange but that will be an overkill

Expected Behavior or Use Case

Presto Component, Service, or Connector

Possible Implementation

Example Screenshots (if appropriate):

Context

[jaystarshot]

cc: @feilong-liu @aditi-pandit

[jaystarshot]

Could be related to #13008