Fix test warnings in dpnp.tensor tests (#2852)

This PR proposes to fix test warnings in `dpnp.tensor` tests by
replacing deprecated strides assignment with
`np.lib.stride_tricks.as_strided` in `test_usm_ndarray_dlpack.py`
and suppressing overflow warnings from np.allclose in
`test_exp.py:test_exp_complex_contig`

Author: vlad-perevezentsev

dpnp/tests/tensor/elementwise/test_exp.py

@@ -78,6 +78,7 @@ def test_exp_real_contig(dtype):
     assert_allclose(dpt.asnumpy(Z), np.repeat(Ynp, n_rep), atol=tol, rtol=tol)
 
 
+@pytest.mark.filterwarnings("ignore:overflow encountered:RuntimeWarning")
 @pytest.mark.parametrize("dtype", ["c8", "c16"])
 def test_exp_complex_contig(dtype):
     q = get_queue_or_skip()

dpnp/tests/tensor/test_usm_ndarray_dlpack.py

@@ -783,7 +783,9 @@ def test_copy_via_host_gh_1789():
     get_queue_or_skip()
     x_np = np.ones((10, 10), dtype="i4")
     # strides are no longer multiple of itemsize
-    x_np.strides = (x_np.strides[0] - 1, x_np.strides[1])
+    x_np = np.lib.stride_tricks.as_strided(
+        x_np, shape=x_np.shape, strides=(x_np.strides[0] - 1, x_np.strides[1])
+    )
     with pytest.raises(BufferError):
         dpt.from_dlpack(x_np)
     with pytest.raises(BufferError):