flip: improve error handling and error messages.

[ysiraichi]

This PR refactors the tensor_methods::flip implementation by improving its error message, and returning a status type value.

Key Changes:

• Make tensor_methods::flip return StatusOr<absl_nonnull XLATensorPtr>
• Improve error message on incompatible tensor shapes

Example:

dims = [0, 0, 0, 1, 2, 3, -1]
a = torch.rand(2, 2, 2, 2, device="xla")
print(torch.flip(a, dims=dims))

Before:

Traceback (most recent call last):
  File "scratch.py", line 8, in <module>
    print(torch.flip(a, dims=dims))
          ^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: Check failed: unique_dims.size() == dimensions.size() (4 vs. 7) (at torch_xla/csrc/tensor_methods.cpp:1675)

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

After:

Traceback (most recent call last):
  File "scratch.py", line 8, in <module>
    print(torch.flip(a, dims=dims))
          ^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: flip(): expected each dimension to appear at most once. Found dimensions: 0 (3 times), 3 (2 times). Consider changing dims from [0, 0, 0, 1, 2, 3, -1] to [0, 1, 2, 3].

Status Propagation Trace:
    From: flip at torch_xla/csrc/tensor_methods.cpp:1704 (error: flip(): expected each dimension to appear at most once. Found dimensions: 0 (3 times), 3 (2 times). Consider changing dims from [0, 0, 0, 1, 2, 3, -1] to [0, 1, 2, 3].)

Exception raised from MaybeThrow at torch_xla/csrc/status.cpp:128 (most recent call first):