Add casting capability to cffi_support.ffi.from_buffer in nopython mode

[dimitar-petrov]

Feature request

Currently casting a numpy array with dtype=np.complex128 and passing it to cffi function is not possible in nopython mode.

from numba import cffi_support, njit
import numpy as np

def f(a):
    cdata = cffi_support.ffi.from_buffer('double []', a)
    
a = np.ascontiguousarray([], dtype=np.complex128)
f(a)
njit(f)(a)

It would be great if different cdata could be passed. Since now it supports only the default cdata="char[]"

TypingError: Failed in nopython mode pipeline (step: nopython frontend)
Invalid use of BoundFunction(ffi.from_buffer for ffi) with parameters (Literal[str](double []), array(complex128, 1d, C))
 * parameterized
[1] During: resolving callee type: BoundFunction(ffi.from_buffer for ffi)
[2] During: typing of call at <ipython-input-27-3dc53d0b7bf1> (4)