Add continuous benchmark

[pradithya]

This PR add continuous benchmark validation and report publishing using https://github.com/benchmark-action/github-action-benchmark.

• Every pull request will be checked against performance regression and will fail if the performance degrade by more than 2x.
• Once a PR is merge to master, a benchmark report will be created and published in https://caraml-dev.github.io/universal-prediction-interface/dev/bench/

[leonlnj]

previously we were benchmarking all the various python version, now we are doing 3.7 only. Is this intended? If so what is the rationale for 3.7?

Left a question, other than that LGTM! Thanks!

[tiopramayudi]

overall is LGTM

[pradithya]

previously we were benchmarking all the various python version, now we are doing 3.7 only. Is this intended? If so what is the rationale for 3.7?

Yeah, turns out it slow down the CI by too much. No particular reason of choosing 3.7, we can pick any version since it compare relative performance.