Light color restoration algorithm produces awful results when using lightdir from lightgrid

[illwieckz]

This use case occurs when the engine is attempting to do normal/specular compute using lightmap light color but lightgrid light direction:

See #299 about lightcolor restoration, #299 is the deluxemap counterpart of this bug.

The bug may also be related to #320.

We must find a better algorithm or a variant of it for that special use case if we want to support it.

We may want to support it for two reasons:

1. reduce the complexity of the code, avoid intrusive hacky ifdefs everywhere.
2. enable normal mapping on legacy maps.

[hypov8]

now that i think back, i previously have a similar issue in kingpinQ3. random artifacts in deluxmap
i found it was a compiler issue.
if you compile with thread 1 it fixed the issue. there is some bug when its laying down the UV map for patches. it would overlap existing data. hence why u see an inverted deluxmap. because it not the only face using the texture space.

to use multi thread still. i just set 1 thread before running
Sys_Printf("--- MapRawLightmap ---\n");
RunThreadsOnIndividual
then set back to previous value