Update dpnp 0.8

[PokhodenkoSA]

• API change starting from use sycl_adapter in krnl_indexing dpnp#902.
• Update to update backend interfaces for dot and matmul functions dpnp#926

@shssf @Alexander-Makaryev @oleksandr-pavlyk
Please, note changes in https://github.com/IntelPython/numba-dppy/pull/524/files#diff-cd1b3e27f060a681f6c297528940f8c3faa8a26f92f009ecc4532b72e61127b8. It is workaround for DPNP issues.

[PokhodenkoSA]

Test fails:

numba_dppy/tests/integration/test_dpnp_interop.py::test_consuming_array_from_dpnp[level_zero:gpu:0-int32] FAILED                                                        [  8%]
numba_dppy/tests/integration/test_dpnp_interop.py::test_consuming_array_from_dpnp[level_zero:gpu:0-int64] FAILED                                                        [  8%]
numba_dppy/tests/integration/test_dpnp_interop.py::test_consuming_array_from_dpnp[level_zero:gpu:0-float32] FAILED                                                      [  8%]
numba_dppy/tests/integration/test_dpnp_interop.py::test_consuming_array_from_dpnp[level_zero:gpu:0-float64] FAILED                                                      [  8%]
numba_dppy/tests/integration/test_dpnp_interop.py::test_consuming_array_from_dpnp[opencl:cpu:0-int32] FAILED                                                            [  8%]
numba_dppy/tests/integration/test_dpnp_interop.py::test_consuming_array_from_dpnp[opencl:cpu:0-int64] FAILED                                                            [  8%]
numba_dppy/tests/integration/test_dpnp_interop.py::test_consuming_array_from_dpnp[opencl:cpu:0-float32] FAILED                                                          [  8%]
numba_dppy/tests/integration/test_dpnp_interop.py::test_consuming_array_from_dpnp[opencl:cpu:0-float64] FAILED   
...
numba_dppy/tests/njit_tests/dpnp/test_numpy_array_creation.py::test_unary_ops[opencl:gpu:0-copy-float64] PASSED                                                         [ 39%]
numba_dppy/tests/njit_tests/dpnp/test_numpy_array_creation.py::test_unary_ops[opencl:gpu:0-trace-int32] Segmentation fault (core dumped)

[PokhodenkoSA]

Fails:

numba_dppy/tests/njit_tests/dpnp/test_numpy_array_ops.py::test_unary_ops[opencl:gpu:0-max-int32-10] Segmentation fault (core dumped)

[PokhodenkoSA]

Fails:

numba_dppy/tests/njit_tests/dpnp/test_numpy_array_ops.py::test_unary_ops[opencl:gpu:0-min-int32-10] Segmentation fault (core dumped)

numba_dppy/tests/njit_tests/dpnp/test_numpy_array_ops.py::test_take[opencl:gpu:0-int32-indices0] Aborted (core dumped)

[PokhodenkoSA]

@shssf @Alexander-Makaryev
I can not merge it because internal CI fails. It fails because we have dpnp 0.7.1=_35 on internal CI.
Also I can not set dependencies dpnp>=0.8* because there are no any tags after 0.7.1.
Please, set tag on your repo.

[Alexander-Makaryev]

@PokhodenkoSA
Ok, I created a tag as you asked.

[Alexander-Makaryev]

@PokhodenkoSA I think that we have misunderstanding about issues that you met.
Please, look at my comment in corresponding topic IntelPython/dpnp#723 (comment)

[PokhodenkoSA]

we have misunderstanding about issues that you met

I agree. I met default behavior of dpnp. I should use DPNP_OUTPUT_DPCTL=1 and unskip tests.

[PokhodenkoSA]

@shssf @Alexander-Makaryev
Please, upload dpnp 0.8 on internal channels.

[Alexander-Makaryev]

@PokhodenkoSA

@shssf @Alexander-Makaryev
Please, upload dpnp 0.8 on internal channels.

As I already said, we will upload new versions after finishing work with previous one. For now it is possible to use ready package from CI or build it from sources.