Type declaration for `indices` and `indptr` might lead to dtype mismatch

[j-i-l]

When passing the arguments of a scipy.sparse.csr_matrix to sparse_stoch_from_full_csr we run into a ValueError:

long long[:] Tf_indices, E ValueError: Buffer dtype mismatch, expected 'long long' but got 'int'

See e.g. here

We can address this by explicitly converting the indices to np.int64 when passing them as arguments, or we might change datat type to int instead of long long in sparse_stochfrom_full_csr (and others).

Note that for some functions in _cython_sparse_stoch.pyx already declare int[:] as data type when indices are passed as arguments.

---

          We have some issues with the type declarations for `indices` and `indptr`:

https://github.com/alexbovet/flow_stability/actions/runs/10513029536/job/29127746857#step:5:116

It also seems not completely consistent in the defs:

E.g. for indptr some functions require int:

https://github.com/alexbovet/flow_stability/blob/8648a8bd715e3b086ed13cb7453d41754c176398/src/flowstab/_cython_sparse_stoch.pyx#L859-L863

some long long: https://github.com/alexbovet/flow_stability/blob/8648a8bd715e3b086ed13cb7453d41754c176398/src/flowstab/_cython_sparse_stoch.pyx#L613-L618

I'd put everything indices and indptr to long and declare with np.int64 on the python site. @alexbovet is that OK for you or is there actually a use-case for long long?

Originally posted by @j-i-l in https://github.com/alexbovet/flow_stability/issues/35#issuecomment-2305337823