[mlir][MemRef] Make fold-memref-alias-ops use memref interfaces

This replaces the large switch-cases and operation-specific patterns
in FoldMemRefAliashops with patterns that use the new
IndexedAccessOpInterface and IndexedMemCopyOpInterface, which will
allow us to remove the memref transforms' dependency on the NVGPU
dialect.

This does also resolve some bugs and potential unsoundnesses:
1. We will no longer fold in expand_shape into vector.load or
vector.transfer_read in cases where that would alter the strides
between dimensions in multi-dimensional loads. For example, if we have
a `vector.load %e[%i, %j, %k] : memref<8x8x9xf32>, vector<2x3xf32>`
where %e is
`expand_shape %m [[0], [1], [2. 3]] : memref<8x8x3x3xf32> to 8x8x9xf32,
we will no longer fold in that shape, since that would change which
value would be read (the previous patterns tried to account for this
but failed).
2. Subviews that have non-unit strides in positions that aren't being
meaningfully accessed (ex. the outer dimensions of a 1-D vector load)
are now folded
3. While it is still not possible to fold a collapse_shape with a
transfer_read or transfer_write if it would affect the transfer
dimensions, the fold will now occur if it would not
4. DMA operations (nvgpu async copy, memref.dma_start) now support
expand_shape and collapse_shape folding in addition to subview.
5. Loading or storing a 1xN vector from a memref where the 1 and N are
the result of expanding the same dimension will now fold into loading
a vector of length N and a shape cast.
6. An issue where IR would be created before pattern failure has been
resolved.

Assisted-By: Claude code (generating copies of similar patterns, test
generation)