Add a16b16g16r16 format
This MR adds an a16b16g16r16
format. This is one of the more common high bit depth formats, as it is equivalent to GL_RGBA16.
This commit only registers the new format with the stress test; most of the others have embedded assumptions about bit depth being <= 32. (If I have time, I'll work on getting those tests to support the higher bit depth formats (rgba_float,rgb_float,any future additions), but that work can be done independently of this MR. Some initial work in this direction is contained in !53.)