It is mapped with the keyboard mapping to the resulting character so
the key 'A' is DIK_A nomatter what its scancode or vkey would be. This
is relevant to Japanese keymapping where the '@' key is in the '['
location the scancode for both is 0x22 but dinput generates DIK_AT in
japanese and DIK_LBRACKET in us_qwerty.

There are also overlay backbuffers, which aren't render targets. The
primarysurface and d3ddevice cap flags should be enough here.

If a format is not supported natively by opengl, a shader may be able
to convert it. Up to now, CheckDeviceFormat had magic knowldge which
GL extensions lead to which supported format. This patch adds
functions that allow CheckDeviceFormat to ask the actual
implementation for its capabilities.