Right now max concurrency as a variable is only used for in-flight requests. Previously it was used for both workers and in-flight requests. We should make it clear and remove any decoupling.

What we need to do

1. Rename max concurrency as a variable for inflight limit
2. Remove max concurrency as a cli arg
3. Bench different values of inflight limit

Motivation:
if max_concurrency cli arg removed, which reduced max_concurrency default from 256 to 32 and this reduction could cause perf degredation. Happened due to how we make use of inflight allowance with workers.

Context:
#18872