Support sparse matrices in cast methods

[stavros11]

Required to enable qibo Hamiltonian creation using sparse matrices. Sparse matrices are supported for both numba (via scipy.sparse) and cupy/cuquantum (via cupy.sparse) platforms.

[codecov]

Codecov Report

Merging #68 (809eb1e) into main (8a1444c) will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##              main       #68   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            9         9           
  Lines         1022      1080   +58     
=========================================
+ Hits          1022      1080   +58     

Flag	Coverage Δ
unittests	100.00% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
src/qibojit/custom_operators/__init__.py	100.00% <100.00%> (ø)
src/qibojit/custom_operators/platforms.py	100.00% <100.00%> (ø)
src/qibojit/tests/test_platforms.py	100.00% <100.00%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8a1444c...809eb1e. Read the comment docs.

[stavros11]

I extended the linear algebra methods for eigenvectors to work with sparse matrices, however it seems that these methods are a bit unstable, particularly for cupy. For example the following script:

from qibo import hamiltonians
from cupy import sparse
from cupy.sparse.linalg import eigsh

h = hamiltonians.TFIM(4, h=1.0)
m = sparse.coo_matrix(h.matrix)

for _ in range(10):
    eigvals, eigvecs = eigsh(m, 6)
    print(eigvals)

gives different values for the eigenvalues of the same matrix in each iteration. This makes testing a bit hard.

[scarrazza]

Thanks, looks good to me.