Remove redundant benchmarks

[mijicd]

Summary

• removed redundant benchmarks
• dropped unnecessary library dependencies

Description

So far, we've been benchmarking individual operations across multiple client libraries. While there was some variance among the absolute values, their orders of magnitude were more or less the same. Instead of proceeding with such an approach, we should construct more specific benchmarks, i.e., ones that exercise critical parts of our code.

Note that the new approach would allow us to run our benchmark suite as part of the CI.