fix(benches): relax `default measurement time`, split `benches` and run `fast call` only on batches (#885)

* fix(benches): relax default and split benches

* decrease slow and memory call overhead

* revert `memory and slow call` + batch on slow_call

* Update benches/helpers.rs

* add asserts for fast_call

* add tokio-console support for benches

* Update examples/examples/tokio_console.rs

* Update examples/examples/tokio_console.rs

* fix(benches): increase sample time for benches

* change to 100s

* split benchmarks slow/fast

* increase default sample time

* Update benches/README.md

* sample_time -> measurment_time

* improve benches with 10 requests per conn

* add subscription bench

* fix nits

* fix nit run 256 too

* add CPU profiling via pprof

* rewrite me

* add move benchmark methods; fast, slow & memory intensive

* add instructions for ulimit

* reduce mem call not reach 256MB limit in WS for batches