Optimise aterm lookups performance

[mlaveaux]

The aterm benchmarks show that aterm lookups are about 3 times more expensive than in the mCRL2 aterm library. This could be inheritedly from the hash table implementation, or the hashing that is being used. In any way we should try to optimised this further. Furthermore, the mCRL2 toolset also has fixed arity constructors that are about 3 times faster than the dynamic ones. See if we can also introduce this using macros.

[mlaveaux]

One big difference in the mCRL2 library are the fixed size arity term constructors. They are in the order of 5 times faster than the current benchmarks. We could experiment with this, and also see if macros could be used to instantiate them inside the rewriter.