Error message exposes internal column name

[vpapavas]

Describe the bug
Error message displayed is:
Invalid key field, not found in schema: KSQL_INTERNAL_COL_1

To Reproduce
Steps to reproduce the behavior, include:

1. The version of KSQL.
   Master
2. Sample source data.

Assume stream with schema
CREATE STREAM metrics_stream (metrics_name STRING, metrics_value DOUBLE, dimensions STRUCT<cluster STRING, topic_name STRING, env STRING>, ts BIGINT) WITH (KAFKA_TOPIC = 'metrics', VALUE_FORMAT = 'JSON', timestamp='ts');

3. Any SQL statements you ran
   create table test as select dimensions->cluster as cluster , collect_set(dimensions->cluster) as collected from metrics_stream_2 group by dimensions->cluster;

Expected behavior
The query should not fail, but even if it were supposed to fail, the error message does not help identifying what is wrong. Moreover, it should not expose the internal column name KSQL_INTERNAL_COL_1.

Actual behaviour
A clear and concise description of what actually happens, including:

1. CLI output
2. Error messages
   Invalid key field, not found in schema: KSQL_INTERNAL_COL_1
3. KSQL logs

Additional context
This query does not fail:
create table test as select metrics_name as cluster , collect_set(metrics_name) from metrics_stream_2 group by metrics_name;
The problem seems to be in the STRUCT column.

[big-andy-coates]

This shouldn't be an issue once ##3537 and then #3669 is done, as 'KEY fields' won't be a thing any more.

[big-andy-coates]

No longer reproducible in master