Fix initialization order for _unit_step_size

[Egor18]

This PR is to fix segmentation fault that occures when align_depth is enabled and we subscribe to one of the camera/aligned_depth_to_* topics.
The issue was that the _unit_step_size was not initialized before getParameters() call, but inside getParameters it is used when calculating _aligned_depth_images vectors size:

realsense/realsense2_camera/src/realsense_node.cpp

Line 286 in f26dbde

_aligned_depth_images[DEPTH].resize(_width[DEPTH] * _height[DEPTH] * _unit_step_size[DEPTH]);

So all the vectors inside _aligned_depth_images had size of 0 and because of that later on data() returned nullptr and we had segfault when trying to call memset with a null pointer

realsense/realsense2_camera/src/realsense_node.cpp

Lines 677 to 680 in f26dbde

	auto& out_vec = _aligned_depth_images[sip];
	alignFrame(_stream_intrinsics[DEPTH], _stream_intrinsics[sip],
	depth_frame, from_image_frame.get_bytes_per_pixel(),
	_depth_to_other_extrinsics[sip], out_vec);

realsense/realsense2_camera/src/realsense_node.cpp

Lines 596 to 599 in f26dbde

	uint8_t* p_out_frame = out_vec.data();
	static const auto blank_color = 0x00;
	memset(p_out_frame, blank_color, other_intrin.height * other_intrin.width * output_image_bytes_per_pixel);