benchmarks: improve code base

Just some refactoring of some benchmarks to reduce code overhead and improve readability.

I some times changed let in a loop to var because it may theoretically still lead to a deopt / prevent a opt.

I fixed two benchmarks that regressed before and what I realized when going over all files again.

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• tests and/or benchmarks are included
• documentation is changed or added
• commit message follows commit guidelines

Affected core subsystem(s)

benchmark