The most complex slots are __new__, __init__ and __call__ because they
accept keywords. It's hard to design a calling convention which optimize
all cases because of the backward compatibility. The risk is to convert a
dict to args + kwnames (tuple) and then back to a dict: two expensive and
useless conversions.

It's my 3rd or 4th attempt to optimize new/init/call :-) My previous
attempt was to add new tp_fastXXX slots to types, but again thr backward
compatibilty was a major pain.
https://bugs.python.org/issue29358

Here the scope is much small and backward compatibilty shouldn't affect us.

I will try to find a way to optimize these slots as well and complete my
patch.