Pull Queries: Support arbitrary pull queries

[big-andy-coates]

We should support a lot more functionality for pull queries, e.g.

• pull queries where the table is not materialized
• pull queries with no constraint on ROWKEY
• pull queries with JOINs or GROUP BY etc.

The distinction here is that a pull query runs from the beginning of time to current offsets. For these more complex queries we should fall back to using a KS topology against the data in Kafka i.e. using materialized state stores is an optimisation.

We can then make enhancements over time to support more and more using materialized tables.

EXPLAIN <some pull query> should make it clear if a query is a 'table scan' ie reading all data from Kafka, or using a materialized table.

[vinothchandar]

@derekjn @apurvam @big-andy-coates Are we planning to work on this in the Q4 timeframe?

Can we prioritize them against feedback we get and schedule them for Q1 instead..

[big-andy-coates]

Really, this requires #3541 first.

[agavra]

Duplicate of #6726