Improved error handling for usm_ndarray.__setitem__

```python
import dpctl.tensor as dpt, numpy as np

u, n = dpt.empty(5), np.ones(5, dtype="O")
u[...] = n
```

This example used to raise a cascade of exceptions starting
with cryptic  "RuntimeError: Unrecogized typenum 17 encountered.",
where typenum 17 corresponds to type object.

After this change, the error message is crisper:

```
ValueError: Input of type <class 'numpy.ndarray'> can not be \
    assigned to usm_ndarray because of unsupported data \
    type 'object'
```

Author: oleksandr-pavlyk

dpctl/tensor/_usmarray.pyx

@@ -1110,13 +1110,19 @@ cdef class usm_ndarray:
                             "converted to usm_ndarray"
                         )
                 else:
+                    rhs_np = np.asarray(rhs)
+                    if type_bytesize(rhs_np.dtype.num) < 0:
+                        raise ValueError(
+                            f"Input of type {type(rhs)} can not be "
+                            "assigned to usm_ndarray because of "
+                            f"unsupported data type '{rhs_np.dtype}'"
+                        )
                     try:
-                        rhs_np = np.asarray(rhs)
                         _copy_from_numpy_into(Xv, rhs_np)
                     except Exception:
                         raise ValueError(
                             f"Input of type {type(rhs)} could not be "
-                            "converted to usm_ndarray"
+                            "copied into dpctl.tensor.usm_ndarray"
                         )
             return