Native TCO

You can actually get native tail call optimisation in Ruby, if this works for Läsp it would perhaps even be nicer than mimicking Clojure's `recur`. It has to be turned on like this:

``` ruby
RubyVM::InstructionSequence.compile_option = { 
  tailcall_optimization: true,
  trace_instruction:     false,
}
```

One pitfall is that this has to be set before the code that uses it is compiled.
