Introduce micro-benchmarks

[Lorak-mmk]

We sometimes want to benchmark some things, like LBP, or ser/deser. We do not have an estabilished way on how to do this. This increases the friction of performing any benchmarks, causing us to perform them very rarely.

I think we should introduce some microbenchmarks that will let us evalute our changes. There currently is one benchmark, a serialization benchmark in scylla-cql, but it doesn't let us easily compare the performance between API versions.
We should have an easy way to compare the performance between revisions.

Related: https://github.com/cvybhu/lbbench LBP benchmark, which we could integrate here.