wineboot: Start an additional 32-bit instance of rundll32 on 64-bit platforms to do the Wow64-specific setup.

Now the shader assembler is structurally complete and it correctly
assembles one trivial shader program.

Currently it only accepts a minimal subset of the syntax (e.g. just an
instruction and two register types supported) and doesn't produce any
real output (i.e. shader bytecode).

This only executes the preprocessing pass. No parsing of the shader yet.

wined3d: Add more general support for partial updates of compressed surfaces to IWineD3DDeviceImpl_UpdateSurface().

wined3d: Use GL_UNPACK_ROW_LENGTH for partial updates of regular surfaces in IWineD3DDeviceImpl_UpdateSurface().

The value is based on 64-bit Win 7, the main thing we care about is that it's
not 0.