Issue #29507: Optimize slots calling Python methods. For Python methods, get
the unbound Python function and prepend arguments with self, rather than
calling the descriptor which creates a temporary PyMethodObject.

Add a new _PyObject_FastCall_Prepend() function used to call the unbound Python
method with self. It avoids the creation of a temporary tuple to pass
positional arguments.

Avoiding temporary PyMethodObject and avoiding temporary tuple makes Python
slots up to 1.46x faster. Microbenchmark on a __getitem__() method implemented
in Python:

Median +- std dev: 121 ns +- 5 ns -> 82.8 ns +- 1.0 ns: 1.46x faster (-31%)
Co-Authored-by: INADA Naoki <songofacandy@gmail.com>