fix[next-dace]: Preserve ITIR data layout

[edopao]

Description

The DaCe backend is reordering the dimensions of field domain based on alphabetical order - we call this the canonical order of dimensions. This PR corrects indexing of field domain in get_scan_dim which was not consistent with the canonical order.

Additional minor edit:

• rename map_domain -> map_ranges
• replace dace.Memlet() with dace.Memlet.simple()

[petiaccja]

Choose dimension ordering and ABI scheme:

• Canonical ordering + no runtime casts: order dimensions lexicographically , shuffle strides, sizes and offsets consistently, everywhere (kinda error prone but simple and no runtime costs)
• Dynamic ordering + runtime casts: generate each callable with whatever dimension order in its ABI, and generate runtime instructions for casting differently ordered arguments to the callable (kinda error prone, kinda complicated, kinda slow, but reuses function bodies)
• Dynamic ordering + no runtime casts: generate each callable once for each different dimension order of its arguments (simple implementation, complicated caching for SDFG generation, not error prone, bloated generated code)

[edopao]

Decided to go for option 1, which means keeping the solution already present in main branch:

• Canonical ordering + no runtime casts: order dimensions lexicographically , shuffle strides, sizes and offsets consistently, everywhere (kinda error prone but simple and no runtime costs)

However, one place needs to be updated because it was not consistent:

• Addressing of scan dimension, which was using the ITIR order of dimensions in the output field domain.