The pickle module was being imported from numpy.core.numeric. It was
defined there in order to use pickle5 when available in Python3 and
cpickle in Python2. The numpy.compat module seems a better place for
that.

This also improves `np.ctypeslib.as_ctypes` to support more types of array:
* non-native endianness
* structured arrays with non-overlapping fields
* booleans

Also switches to using a parametrized test, for better error messages

- https://bugs.python.org/issue35507
- https://stackoverflow.com/q/53757856/812183

* Review F401,F841,F842 flake8 errors (unused variables, imports)

* Review comments

* More tests in test_installed_npymath_ini

* Review comments

This:
* fixes a regression in 1.15, where it became impossible to set the return value of a cdll function to an ndpointer.
* removes ndpointer.__array_interface__, which was being ignored anyway in favor of the PEP3118 buffer protocol
* adds `ndpointer.contents` to recover the lost functionality, while staying in line with the ctypes behavior
* removes another instance of `descr`, which enables overlapping fields to be returned from C functions (such as unions).
* Fixes a long-term bug where using ndpointer as a return type without specifying both type and dtype would produce an object array containing a single `ndpointer`. Now the ndpointer is returned directly.

This relates to gh-12421, and likely fixes toinsson/pyrealsense#82

Fixes an alarming bug introduced in gh-7311 (1.12) where the following is true

    np.ctypeslib.ndpointer(ndim=2) is np.ctypeslib.ndpointer(shape=2)

Rework of gh-11536