A small test package to demonstrate ILP64 support in downstream Cython packages.

Currently, SciPy can be built with or without ILP64 support; in the former case 
`cython_lapack` ABI can be built to be either LP64 or ILP64, see [1] for a discussion.

Here we demonstrate two ways of adapting downstream for a Cython-using package,
which relies on `cython_lapack`.

A difficulty is that `cython_lapack` is using `blas_int` type for integer variables
and arrays, which resolves to either a C `int` or `int64_t`.
A downstream user package has two options:

1. Use a cython wrapper to cast from `int` to `blas_int` type, so that other modules
   can depend on this wrapper and hence don't need changing.
   This way, the linear algebra is exposed as LP64 regardless of whether SciPy
   is built to internally use LP64 or ILP64 variants.
   See `tests/test_wrapper.py` for details.

2. Directly use SciPy's `blas_int` type without an intermediate wrapper layer.
   This way, the downstream functionality is not limited by `INT_MAX` for array sizes.
   (see `tests/test_direct.py` for details).

See the relevant SciPy CI workflow [2] for an example of building and testing
this package.

------------

[1] https://scipy.github.io/devdocs/building/blas_lapack.html#bit-integer-ilp64-blas-lapack
[2] https://github.com/scipy/scipy/blob/main/.github/workflows/linux_blas.yml

