Use masked_fill_ in replace_masked_values

[matt-gardner]

I wrote this replace_masked_values function before I realized that masked_fill_ exists. I'm pretty sure we could just use masked_fill_ instead of what's currently here:

allennlp/allennlp/nn/util.py

Lines 463 to 475 in 1700db4

	def replace_masked_values(tensor: Variable, mask: Variable, replace_with: float) -> Variable:
	"""
	Replaces all masked values in ``tensor`` with ``replace_with``. ``mask`` must be broadcastable
	to the same shape as ``tensor``. We require that ``tensor.dim() == mask.dim()``, as otherwise we
	won't know which dimensions of the mask to unsqueeze.
	"""
	# We'll build a tensor of the same shape as `tensor`, zero out masked values, then add back in
	# the `replace_with` value.
	if tensor.dim() != mask.dim():
	raise ConfigurationError("tensor.dim() (%d) != mask.dim() (%d)" % (tensor.dim(), mask.dim()))
	one_minus_mask = 1.0 - mask
	values_to_add = replace_with * one_minus_mask
	return tensor * mask + values_to_add

I'm not certain this works correctly with gradients and things, but it should...