Add benchmarking

[fikovnik]

We need benchmarks.

Right now, we have some benchmarks from RBenchmarking which are run alongside the R BC interpreter.
They are not great - they have to be modified so all functions are compiled, plus they run in the same session.

I think it will be better to support the RBenchmarking directly.
What do we want to measure:

• peak performance -- when the function is compiled into native code
• infrastructure overhead
  • with hot and cold cache

We want to compare with:

• GNU-R BC
• GNU-R AST
• RIR?
• FastR?