scheduler filter include streaming part

[szuecs]

Is your feature request related to a problem? Please describe.

As owner of a streaming application that only support N concurrent users I want to be able to limit the traffic to N.
Right now if you do have a backend that does async response where the streaming part can be long, you get much more than N requests, because the scheduler filters response part is triggered before the response body streaming is happening.

Describe the solution you would like

Make it possible to limit streaming concurrency, too.
Basically decrease the concurrency counter after body streaming was done.

Describe alternatives you've considered (optional)

We tried lifo which does not work as expected and a workaround seems to be backendRatelimit which has a lower error margin than lifo for an application which streams the response.

Additional context (optional)
internal finding