Don't require AS_VALUE to access multi-column group by columns

[agavra]

ksqlDB should allow access to the group-by columns without using AS_VALUE to manually copy them into the value.

Current Behavior:

ksql> CREATE STREAM a (id int key, col1 int) with (kafka_topic='a', value_format='json', partitions=1);
ksql> CREATE TABLE gb AS SELECT id, col1, COUNT(*) FROM a GROUP BY id, col1;

 Message
----------------------------------
 Created query with ID CTAS_GB_15
----------------------------------
ksql> SELECT id FROM gb;
Line: 1, Col: 8: SELECT column 'ID' cannot be resolved.

There are various ways to implement this:

1. we can actively copy any selected group by component into the value
2. we can leverage structured keys (Add support for generic message keys (e.g. message keys in Avro or JSON format, STRUCT) #824) and decompose the key from the struct

I think the former is a good enough solution in the short term (which is how we implemented it before), while we should strive to implement the latter. The latter may require a refactor (TODO: link ticket here).