ggml : update ggml_backend_cpu_device_supports_op

[ggerganov]

This fixes the following failures:

./build/bin/test-backend-ops -b CPU -o CPY

  CPY(type_src=f16,type_dst=q6_K,ne=[256,4,4,4],permute=[0,0,0,0]): OK
  CPY(type_src=f16,type_dst=q6_K,ne=[256,2,3,4],permute=[0,2,1,3]): OK
  CPY(type_src=f16,type_dst=iq2_xxs,ne=[256,4,4,4],permute=[0,0,0,0]): not supported [CPU] not supported [CPU] 
  CPY(type_src=f16,type_dst=iq2_xxs,ne=[256,2,3,4],permute=[0,2,1,3]): not supported [CPU] not supported [CPU] 
  CPY(type_src=f16,type_dst=iq2_xs,ne=[256,4,4,4],permute=[0,0,0,0]): not supported [CPU] not supported [CPU] 
  CPY(type_src=f16,type_dst=iq2_xs,ne=[256,2,3,4],permute=[0,2,1,3]): not supported [CPU] not supported [CPU] 
  CPY(type_src=f16,type_dst=iq2_s,ne=[256,4,4,4],permute=[0,0,0,0]): /llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c:2996: fatal error
/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c:2996: /llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c:2996: fatal error
/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c:2996: fatal errorfatal error/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c:2996: 
/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c:2996: fatal error/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c:2996: fatal error
/llama.cpp/ggml/src/ggml-cpu/ggml-cpu.c:2996: fatal errorAbort trap: 6

[ggerganov]

@slaren Is this a valid fix, or is it better to skip these types in the tests?

[slaren]

My understanding is that these types are effectively unusable without an imatrix, so they should be entirely disabled for quantizing in ggml_cpy. I think what needs to be fixed is ggml_backend_cpu_device_supports_op.