Fix VFA validation for NumPy flip angles

[ParikshitGupta2605]

Summary

This PR fixes the validation of flip_angles in compute_t1_vfa() when provided as a NumPy array.

Previously, the code used:

if not params.flip_angles:

which raises a ValueError for NumPy arrays because their truth value is ambiguous.

Changes

• Replace the boolean truth-value check with explicit validation for None and empty arrays.
• Convert flip_angles using np.asarray() before validation.
• Add a regression test to ensure NumPy arrays are accepted.

Fixes #159

[ltorres6]

@MohamedNasser8 can you review this one when you have some time?

[MohamedNasser8]

Of course

[MohamedNasser8]

Thanks @ParikshitGupta2605

My one concern is where to fix it. This solves it inside vfa.py, but flip_angles actually belongs to DCEAcquisitionParams, and other parts of the code read it too (e.g. signal-to-concentration) . The field is declared as list[float] but nothing actually enforces that. That means every other function depends on DCEAcquisitionParams keeps the same bug.

I think it's better to fix it once at the dataclass for example use post_init to keep it list. I mean convert whatever is passed into a plain list of floats, and raise a clear error if it genuinely can't be converted.

[ParikshitGupta2605]

Thanks for the feedback! Normalizing flip_angles in DCEAcquisitionParams.post_init is a better approach since it guarantees a consistent list[float] representation for all consumers, rather than handling it only in vfa.py. I'll update the implementation accordingly and simplify the VFA-specific changes.

[ltorres6]

Thanks both. Also note there are several locations where we cast as list or as array that probably needs to be cleaned up once this is implemented in the dataclass.

[ParikshitGupta2605]

Reopening with the final version - Pull request #167