Build npairs on windows

[seanpmorgan]

Npairs can not run on windows due to no OpKernel registered SparseMatMul. This appears to have a related issue tensorflow/tensorflow#22699 which has no resolution.

tensorflow.python.framework.errors_impl.InvalidArgumentError: No OpKernel was registered to support Op 'SparseMatMul' used by {{node MatMul}}with these attrs: [transpose_b=true, Ta=DT_FLOAT, Tb=DT_FLOAT, b_is_sparse=true, a_is_sparse=true, transpose_a=false]
Registered devices: [CPU]
Registered kernels:
  <no registered kernels>

         [[MatMul]]
         [[npairs_multilabel_loss/PartitionedCall]]

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "C:\Users\SeanM\Miniconda3\lib\site-packages\tensorflow_core\python\framework\test_util.py", line 1111, in decorated
    f(self, *args, **kwargs)
  File "\\?\C:\Users\SeanM\AppData\Local\Temp\Bazel.runfiles__8x3qnhl\runfiles\__main__\tensorflow_addons\losses\npairs_test.py", line 89, in test_single_label
    self.assertAllClose(loss, 0.253856)
  File "C:\Users\SeanM\Miniconda3\lib\site-packages\tensorflow_core\python\framework\test_util.py", line 1153, in decorated
    return f(*args, **kwds)
  File "C:\Users\SeanM\Miniconda3\lib\site-packages\tensorflow_core\python\framework\test_util.py", line 2495, in assertAllClose
    self._assertAllCloseRecursive(a, b, rtol=rtol, atol=atol, msg=msg)
  File "C:\Users\SeanM\Miniconda3\lib\site-packages\tensorflow_core\python\framework\test_util.py", line 2462, in _assertAllCloseRecursive
    (path_str, path_str, msg)))
  File "C:\Users\SeanM\Miniconda3\lib\site-packages\tensorflow_core\python\framework\test_util.py", line 2355, in _assertArrayLikeAllClose
    a = self._GetNdArray(a)
  File "C:\Users\SeanM\Miniconda3\lib\site-packages\tensorflow_core\python\framework\test_util.py", line 2349, in _GetNdArray
    a = self.evaluate(a)
  File "C:\Users\SeanM\Miniconda3\lib\site-packages\tensorflow_core\python\framework\test_util.py", line 2061, in evaluate
    return sess.run(tensors)
  File "C:\Users\SeanM\Miniconda3\lib\site-packages\tensorflow_core\python\framework\test_util.py", line 1638, in run
    return super(ErrorLoggingSession, self).run(*args, **kwargs)
  File "C:\Users\SeanM\Miniconda3\lib\site-packages\tensorflow_core\python\client\session.py", line 960, in run
    run_metadata_ptr)
  File "C:\Users\SeanM\Miniconda3\lib\site-packages\tensorflow_core\python\client\session.py", line 1183, in _run
    feed_dict_tensor, options, run_metadata)
  File "C:\Users\SeanM\Miniconda3\lib\site-packages\tensorflow_core\python\client\session.py", line 1361, in _do_run
    run_metadata)
  File "C:\Users\SeanM\Miniconda3\lib\site-packages\tensorflow_core\python\client\session.py", line 1386, in _do_call
    raise type(e)(node_def, op, message)
tensorflow.python.framework.errors_impl.InvalidArgumentError: No OpKernel was registered to support Op 'SparseMatMul' used by node MatMul (defined at \\?\C:\Users\SeanM\AppData\Local\Temp\Bazel.runfiles__8x3qnhl\runfiles\__main__\tensorflow_addons\losses\npairs_test.py:78) with these attrs: [transpose_b=true, Ta=DT_FLOAT, Tb=DT_FLOAT, b_is_sparse=true, a_is_sparse=true, transpose_a=false]
Registered devices: [CPU]
Registered kernels:
  <no registered kernels>

         [[MatMul]]
         [[npairs_multilabel_loss/PartitionedCall]]

[seanpmorgan]

This appears to be fixed for TF2.2. Waiting for master to switch to >= 2.2 and then this can be closed.