By analogy with smp_groups which is already public we need to
access this outside of application.cc to pass the right scheduling
groups through to instantiations of services in our redpanda fixture.

This scheduling group is used for consumer fetch processing. We assign
it the same priority as the default group (where most other kafka
handling takes place), but by putting it into its own group we prevent
non-fetch requests from being significantly delayed when fetch requests
use all the CPU.

Issue redpanda-data/core-internal#435

Add the with_scheduling_group call to switch to the kafka scheduling
group in the fetch handling flow. See prior change for context.

Issue redpanda-data/core-internal/redpanda-data#435.

Introduce a config tunable which would let users set the scheduling
group to use for kafka fetch request back to the default group.

Also useful for quick A/B testing of the perf effect of the change.