Do not add Concat + Reshape in OneHotEncoder converter

[DeltaEpsilon7787]

microsoft/onnxruntime#18918

This issue was discovered because sklearn2onnx always adds a Concat + Reshape operator after a OneHotEncoder, which does nothing but waste time for the output there.

Consider removing

sklearn-onnx/skl2onnx/operator_converters/one_hot_encoder.py

Lines 186 to 205 in fdb52ce

	concat_result_name = scope.get_unique_variable_name("concat_result")
	apply_concat(scope, result, concat_result_name, container, axis=-1)
	reshape_input = concat_result_name
	if np.issubdtype(ohe_op.dtype, np.signedinteger):
	reshape_input = scope.get_unique_variable_name("cast")
	apply_cast(
	scope,
	concat_result_name,
	reshape_input,
	container,
	to=onnx_proto.TensorProto.INT64,
	)
	apply_reshape(
	scope,
	reshape_input,
	operator.output_full_names,
	container,
	desired_shape=(-1, categories_len),
	)