[RFC] a benchmark suite for cloudpickle.

FYI, I created a benchmark suite for `cloudpickle` here: [cloudpickle_benchmarks](https://github.com/pierreglaser/cloudpickle_benchmarks).

This should make us gain visibility when working on performance improvements.
I'm thinking we could run a benchmark comparison with `master` for each PR if a commit message contains a special pattern such as `ci run-benchmarks`.

Also, we can include this benchmark repository to @tomaugspurger's machine, that continuously runs benchmark for many `PyData` projects such as `pandas`, `scikit-learn`, `joblib` and so on.