random_: improve error handling and error messages.

[ysiraichi]

This PR refactors the random_ operation implementation by improving its error message, and returning a status type value.

Key Changes:

• Make tensor_methods::random_ return Status
• Replace CheckRangeValues by CheckValueWithinTypeRange, and make it return Status
• Refactor XLANativeFunctions::random_ overloads to handle the status values
• Improve error messages

Example:

tensor = torch.empty(10, device=torch_xla.device())
from_ = 3
to_ = 1
tensor.random_(from_, to_)

Before:

Traceback (most recent call last):
  File "example.py", line 7, in <module>
    tensor.random_(from_, to_)
RuntimeError: Check failed: from <= to_val (3 vs. 1) (at torch_xla/csrc/aten_xla_type.cpp:3031)

Exception raised from operator& at torch_xla/csrc/runtime/tf_logging.cpp:26 (most recent call first):

After:

Traceback (most recent call last):
  File "example.py", line 7, in <module>
    tensor.random_(from_, to_)
RuntimeError: random_(): expected `from` (3) to be smaller than `to` (1).

Status Propagation Trace:
    From: random_ at torch_xla/csrc/tensor_methods.cpp:2870 (error: random_(): expected `from` (3) to be smaller than `to` (1).)

Exception raised from OkOrThrow at torch_xla/csrc/status.cpp:121 (most recent call first):

[zhanyong-wan]

Thanks!