Fixed dpt.arange(9.7, stop=200, step=100, dtype='i4')

To align with behavior of numpy, changed the logic of determining
step argument for the call to `_linspace_step` routine. We now
compute first and second element of the array of given type, and
determine the step as a the difference of these.

To avoid possible overflow message when subtracting unsigned integers,
cast first and second element to int64, subtract, and cast to the
target type.

```ipython
In [1]: import dpctl, dpctl.tensor as dpt, numpy as np

In [2]: dpt.asnumpy(dpt.arange(9.7, stop=200, step=100, dtype='i4'))
Out[2]: array([  9, 109], dtype=int32)

In [3]: np.arange(9.7, stop=200, step=100, dtype='i4')
Out[3]: array([  9, 109], dtype=int32)
```

Author: oleksandr-pavlyk

dpctl/tensor/_ctors.py

@@ -569,9 +569,18 @@ def arange(
         buffer_ctor_kwargs={"queue": sycl_queue},
     )
     sc_ty = dt.type
-    _step = sc_ty(start + step) - start
-    _step = sc_ty(_step)
-    _start = sc_ty(start)
+    _first = sc_ty(start)
+    if sh > 1:
+        _second = sc_ty(start + step)
+        if dt in [dpt.uint8, dpt.uint16, dpt.uint32, dpt.uint64]:
+            int64_ty = dpt.int64.type
+            _step = int64_ty(_second) - int64_ty(_first)
+        else:
+            _step = _second - _first
+        _step = sc_ty(_step)
+    else:
+        _step = sc_ty(1)
+    _start = _first
     hev, _ = ti._linspace_step(_start, _step, res, sycl_queue)
     hev.wait()
     if is_bool: