fix vck5000 tests (again)

[fifield]

2nd try.

[fifield]

The lowering to vck5000 (in AIETargetHSA) seems to depend on NpuDmaMemcpyNdOp, which I'm not sure is the case any longer in the failing examples. I haven't run locally to learn all the details. There may be more failures masked by REQUIRES peano (i.e. they don't run).

2024-12-18T18:29:56.3638606Z /mlir-aie/programming_examples/basic/matrix_scalar_add/test_vck5000.cpp:113:3: error: no matching function for call to 'invoke_data_movement'
2024-12-18T18:29:56.3639240Z   invoke_data_movement(queues[0], &agents[0], in_a, out);
2024-12-18T18:29:56.3639516Z   ^~~~~~~~~~~~~~~~~~~~
2024-12-18T18:29:56.3640156Z /mlir-aie/build/programming_examples/basic/matrix_scalar_add/aie.mlir.prj/aie_data_movement.cpp:10:6: note: candidate function not viable: requires 2 arguments, but 4 were provided
2024-12-18T18:29:56.3641131Z void invoke_data_movement(hsa_queue_t *q, hsa_agent_t *a) {
2024-12-18T18:29:56.3641428Z      ^
2024-12-18T18:29:56.3641584Z 1 error generated.

In any case the failure makes sense if there are no longer any memcpynd in the generated runtime sequence.

cc: @eddierichter-amd

[fifield]

Based on discussion I've XFAIL'd the tests for now.

[fifield]

Nth time is the charm?
https://github.com/Xilinx/mlir-aie/actions/runs/12399512776
https://github.com/Xilinx/mlir-aie/actions/runs/12400558087