Adding support for numpy 2

[prsabahrami]

Adhering to NEP 50, np.can_cast no longer applies any value-based logic for 0-D arrays and NumPy scalars.

Using np.min_scalar_type(value), the data type with the smallest size and smallest scalar kind that can hold value is checked against dtype.

However, there is still another issue due to the below schema of numpy 2.0.0's promotion rules:

For example, in the case of casting np.int64(5) to np.int8 , np.min_scalar_type(np.int64(5)) returns np.uint8, and according to the diagram, np.uint8 can not be casted to np.int8. I have manually checked if value is of kind uint and dtype is of kind np.signedinteger and if the casting can safely happen.

[pavelzw]

Thanks!

[prsabahrami]

My pleasure!