[ci] Additional variables for benchmarks execution

[alvicsam]

Added MEASUREMENT_TIME: 3 and SLOW_MEASUREMENT_TIME: 40 variables to benchmark job. With these variables it takes ~2h to run benchmarks: https://gitlab.parity.io/parity/mirrors/jsonrpsee/-/jobs/1885548

cc https://github.com/paritytech/ci_cd/issues/302

[niklasad1]

I suggest to make slow stuff 60s but I have done some investigation and with some improvement it still took ~3h to complete.

I need to refactor the benches to be faster but the bad the thing is that we can't compare with the old data then :(

See #885

I still think 3h is way too long.

[alvicsam]

I need to refactor the benches to be faster but the bad the thing is that we can't compare with the old data then :(

If we consider some data not accurate maybe it's not that bad?

I still think 3h is way too long.

But why? Because for local run it's too long?

[niklasad1]

If we consider some data not accurate maybe it's not that bad?

Probably yes, but I can try backporting the benches to the "most recent" releases to have some data compare against at least.

But why? Because for local run it's too long?

Yeah or unnecessary load on the benchmarks itself, we are testing with some extreme limits such a calls with 1 MB payload, slow call with a sleep for 1 ms and up to 1024 concurrent requests.

I think we can relax the payload size and sleep time to make it run "reasonable time (less than 1 hour)" without any impact. Trying it right now locally...

[alvicsam]

Also it's possible to run benchmarks with cargo-nextest after criterion support is added to nextest: nextest-rs/nextest#96. It can significantly reduce time

[niklasad1]

I just merged #885 but I think we should try with MEASUREMENT_TIME=10 to begin with it's just used a two benches now.

[alvicsam]

Okay, I'd like to wait https://gitlab.parity.io/parity/mirrors/jsonrpsee/-/pipelines/217578 to check how long the benches will take with this settings. If the job succeeds I'll merge this PR