Use cpdef instead of a wrapper "py_<function>" function where appropriate

[sclamons]

Bioscrape uses a lot of cdef'ed functions with a wrapper for python, e.g. the cdef function "simulate_model" (which cannot be called from python) wrapped with a python function "py_simulate_model". This is a little clunky, and mysterious to users who aren't familiar with Cython.

Cython has a built-in keyword "cpdef" that makes a function both typed and python-callable. Calling a cpdef function is slower than calling a cdef function (see https://notes-on-cython.readthedocs.io/en/latest/fibo_speed.html), but that only affects the time required to call the function, not to run its contents (see https://stackoverflow.com/questions/49172528/should-i-define-my-cython-function-using-def-cdef-or-cpdef-for-optimal-perform). For functions that only get called a small number of times, like simulate_model, we should use cpdef instead of cdef so we don't have the weird "py_" handle gunking up our function names.

This should be a quick fix, but someone will need to think about which functions need "py_" wrappers (for speed) and which can be cpdef-ed safely (for comprehensibility and convenience).

[WilliamIX]

The simulate functions are actually called fairly frequently, especially with interacting lineages. In addition, simulate_single_cell is written in such a way as to reuse memory buffers, which have to be initializes separately. This is very important for performance.

I also think that having py_simulate functions which are not connected to classes is helpful so we can use Bioscrape for scripting without instantiating the classes.

I understand your confusion though. One possible thought is to move all the py_simulate functions into a separate file called wrappers. Would this help make things clearer?