Querier worker not using correct parallelism when using match_max_concurrent

[slim-bean]

Describe the bug
We have found that when using the setting match_max_concurrent our queriers are only processing 1 query in parallel instead of the expected queriers max_concurrent setting. This can be seen by looking at the loki_inflight_requests metric on the queriers which should show a busy querier dong max_concurrent inflight requests.

As a workaround, in the worker config section you can remove the match_max_concurrent setting and instead set parallelism equal to max_concurrent

This was introduced in #4301, first affected released would be 2.4