Use the BenchmarkRunner reactor in all C benchmarks

[cmnrd]

Some benchmarks are significantly influenced by caching effects. We observed this in particular in the Recursive Matrix Multiplication benchmark (see #749). By using the benchmark runner reactor, we can ensure that the benchmarks run with a warm cache. For consistency, and also to integrate the updated version of MatMul.lf with our runner script, we should update all C benchmarks to use the BenchmarkRunner reactor. Note that this BenchmarRunner should use the simplified pattern currently implemented in C++. The version currently used in MatMul.lf is unnecessarily complex.