More filesystem benchmarks

[phischu]

@marvinborner I would like these and the other benchmarks in input_output integrated into effekt plots.

[b-studios]

This will be problematic when run in parallel by the different backend tests, like before.

[marvinborner]

The current setup runs every benchmark that's listed in benchmarks/config_${BACKEND}.txt with their respective argument. Since the IO benchmarks don't use a configurable argument, we can just add the filename without it -- the benchmark will be run regardless.

We can add the lines in this PR, do we want to run them for JS and LLVM?

(edit: I've since realized that IO benchmarks also use arguments)

[phischu]

For larger input sizes many of these segfault on the llvm backend. I am wondering why...

[phischu]

I figured it out: classical buffer overflow