Improve non-uniform 3D ray cast selection

[afritz1]

See #164 and #169.

3D ray cast selection is implemented but it is not completely correct.

Visual debugging code is available at the highest profiler level. Rays are cast out from the camera and draw dots on the screen corresponding to the Y voxel coordinate of the hit. Red dots are Y=0, green are Y=1, etc.. Ideally the dots would perfectly match the visual geometry they are associated with.

Places in code:

OpenTESArena/OpenTESArena/src/Game/Physics.cpp

Line 999 in ac8b298

// @todo: this is very likely the cause of ceilingHeight looking wrong (too far above and below Y=1)

OpenTESArena/OpenTESArena/src/Interface/GameWorldPanel.cpp

Line 106 in ac8b298

void DEBUG_ColorRaycastPixel(Game &game, Renderer &renderer)

Other things of note:

• Seems to look fine when ceilingHeight > 1.0?
• Sort voxelEntityMap pair.value by projected entity position Z
• Only check successful door voxel hits if not in LevelData::openDoors vector
• The world is stretched vertically 20% by SoftwareRenderer::TALL_PIXEL_RATIO

In the picture, the ground should be completely full of red dots, but the ray cast algorithm doesn't register hits everywhere due to bugs in the math calculations, probably deltaDist.

[afritz1]

Fixed in commit da5b86f.

Leaving this open until these are done:

• Sort voxelEntityMap pair.value by projected entity position Z (can check with trees that overlap each other at certain angles; slightly separate problem: columns that go through the ceiling in the Halls of Colossus)
• Only check successful door voxel hits if not in LevelData::openDoors vector

[afritz1]

The initial ray cast bug seems to still appear on the underside of raised platforms in the wilderness as of current master. Could be because the raised platforms extend outside the voxel they live in. Still need to fix raised platform Y offset and size calculation in wilderness first.

[afritz1]

Ray cast selection still seems incorrect, i.e. trying to select a voxel face doesn't always work, so this issue is still relevant.

[afritz1]

Made a few significant bug fixes for selecting voxels across chunk boundaries, but there still feel like a couple minor edge cases with the math when selecting diagonally, like sometimes it's easier to select the farther part of a voxel than the nearer part.

The debug pixel coloring seems all correct now, except for the part where SDL 2.0.10 changed the coloring to be batched, so only one color shows per frame :/