Refactor Datastore to use scipp.DataGroup

[AndrewSazonov]

The current Datastore implementation is based on numpy arrays. With the planned move to scipp, we should refactor Datastore to use scipp.DataGroup (and DataArray where appropriate).

Benefits:

• Native support for uncertainties and masks (useful for excluded regions)
• Cleaner grouping of related arrays (measured, calculated, background, etc.)
• Better consistency with other EasyScience components that are moving to scipp

This was touched on in #88 (review), but needs a dedicated discussion and design decision.

Open questions:

• Should we map each array (e.g., measured intensities) directly to DataArray entries in the DataGroup?
• Should recalculated fields (e.g., calculated intensities, background, d_spacing, etc.) also be stored in the DataGroup, or always computed on demand?
• Is the overall datastore.py structure (BaseDatastore, PowderDatastore, SingleCrystalDatastore) still sufficient, or do we anticipate larger design changes?

[easyscience]

Added default label [priority] ⚠️ label needed. Please consider re-labeling this issue appropriately.