Fix missing support of compute follows data in dpnp.erf

[antonwolfy]

A missing part of compute follows data implementation was added for dpnp.erf function.

Tests for dpnp.erf validation was included into a scope of checks triggered on both Linux and Windows platforms by github action. It's enough to secure the fix is working, since the tests running on CPU device with the latest DPCtl from dppy/label/dev channel reproduce the same fault as reported in #1192.

• Have you provided a meaningful PR description?
• Have you added a test, reproducer or referred to issue with a reproducer?
• Have you tested your changes locally for CPU and GPU devices?
• Have you made sure that new changes do not introduce compiler warnings?
• If this PR is a work in progress, are you filing the PR as a draft?

[oleksandr-pavlyk]

@DenisScherbakov Were you able to confirm that this fix addresses the issue with dpbench?

[DenisScherbakov]

@oleksandr-pavlyk , see below results:

python -c "import dpbench; dpbench.run_benchmark("black_scholes")"

================ Benchmark black_scholes ========================

OMP: Info #277: omp_set_nested routine deprecated, please use omp_set_max_active_levels instead.
/.../envs/intelpython_dpbench/lib/python3.9/site-packages/dpbench/benchmarks/black_scholes/black_scholes_numpy.py:8: RuntimeWarning: divide by zero encountered in true_divide
invsqrt = lambda x: np.true_divide(1.0, np.sqrt(x))
================ implementation dpnp ========================
implementation: dpnp
framework: dpnp
framework version: IntelLLVM 2022.2.0
setup time: 140693690
warmup time: 417284547
teardown time: 1148753
max execution times: 332191359.0
min execution times: 166785354.0
median execution times: 253783139.0
repeats: 10
preset: S
validated: 0

[oleksandr-pavlyk]

Confirmed:

>>> x = dpnp.ones(10**5, 'd')
>>> with dpctl.utils.onetrace_enabled():
...    y = dpnp.erf(x)
...
Device Timeline (queue: 0x55f9b8e3ae20): zeCommandListAppendMemoryCopy(M2S)[8 bytes]<7.1> [ns] = 104752417288 (append) 104752940683 (submit) 104753022849 (start) 104753031515 (end)
Device Timeline (queue: 0x55f9b8e3ae20): zeCommandListAppendMemoryCopy(M2S)[8 bytes]<8.1> [ns] = 104758769736 (append) 104758943047 (submit) 104759023713 (start) 104759032379 (end)
Device Timeline (queue: 0x55f9b8e3ae20): zeCommandListAppendMemoryCopy(M2S)[8 bytes]<9.1> [ns] = 104769063726 (append) 104769220997 (submit) 104769282163 (start) 104769290829 (end)
Device Timeline (queue: 0x55f9b8e3ae20): _ZTSZZN6oneapi3mkl2vm6detail3gpu5intel3usm6linear3v11ILb0EddNS4_7functor12erf_d_ha_genEEEN2cl4sycl5eventERKNS2_5input3usm6linear3v11IT0_T1_EENS2_11tuning_infoERKT2_ENKUlRNSB_7handlerEE_clESR_EUlNSB_2idILi1EEEE_[SIMD32 {3125; 1; 1} {32; 1; 1}]<10.1> [ns] = 104769466295 (append) 104769476408 (submit) 104769797408 (start) 104769940574 (end)
>>> x.sycl_queue == y.sycl_queue
True