[bug] fix xavier uniform init for output layers

[hjlee1371]

In TransformerConfig, initializations for the output layers should be specified as output_layer_init_method.

Megatron-LM/megatron/core/transformer/transformer_config.py

Lines 110 to 113 in db3a3f7

	output_layer_init_method: Callable = None
	"""Method to initialize weights of the output layer of both attention and MLP blocks. If None,
	will be set to megatron.core.utils.scaled_init_method_normal(init_method_std) which is torch nn
	init normal with mean=0.0 and std=init_method_std / math.sqrt(2.0 * num_layers)."""

But Xavier uniform initialization with --init-method-xavier-uniform utilizes scaled_init_method instead. This PR addresses the bugs encountered when utilizing Xavier uniform initialization.