Turns out the original fix was correct for fixed function, but for the
wrong reason. The shader path was already correct. This fixes a
regresssion introduced by 932e95c1.

It seems that with Win2k3, the meaning of SO_REUSEADDR got closer to
BSD semantics. This is nice, as it allows us to finally work like some
version of Windows here.

Currently, stdole2.tlb isn't generated correctly and causes "<failed>"
to appear instead of "FontEvents" in the IFontEventsDisp typedef when
viewed with oleview.

The problem is that the typedef should just generate a VT_USERDEFINED
record, but ends up generating a VT_PTR -> VT_USERDEFINED. So remove
the extra writing of VT_PTR entries.

Fix the skipped pointers checks in encode_type and encode_type to
specifically detect VT_UNKNOWN and VT_DISPATCH types which don't need
one level of pointers, whereas interfaces encoded as VT_USERDEFINED
do.