Ascadian Isle Trees, reworked with LODs and corrected UVs

[BethOfDeath]

flora_tree_04:
- Booleaned tree trunks to get rid of clipping calculations.
- Updated and simplified collision mesh. Previous MOP version 88 tris, now 48 tris.
- Adjusted vertices on leaf meshes to hopefully reduce self-intersection without creating tiny visble gaps in the transparent model.
- Removed identical solid-white vertex colors from leaves.
- Created two LODs with simplified visual geometry, and merged many duplicates.
- Normals work, using G7's advice on leaves.
- Vanilla: 25155 bytes; previous MOP version: 24269 bytes; now: 30588 bytes. (22% increase in filesize over vanilla, 26% increase over previous MOP)

flora_tree_01:
- Simplified and split collision mesh. Was 103 tris, now 60.
- Adjusted LOD switchover distance to occur sooner.
- Removed solid #ffffffff vcols from both sets of leaves and cleared bogus sharp edges.
- Fixed overlapping faces of leaves from vanilla's apparent naive duplication-flip.
- Fixed clipping of part of leaf texture. Entire leaf now visible.
- Previous MOP version used duplicate NiTriShapeData for LOD1 leaves. Now using link to LOD0's identical version.
- Normals work, using G7's advice on leaves.
- LOD0 620 tris. LOD1 474. LOD2 232.
- Vanilla was 66737 bytes, previous MOP version 95686 bytes, now 62336 bytes. (7% reduction from vanilla, 35% from previous MOP.)

flora_tree_02:
- Simplified collision mesh. Corrected apparently accidental inclusion of a couple faces when splitting collision mesh, resulting in nearly as large a calculation volume as an unsplit mesh, in the process. -- Now only 3 pieces instead of 5, as the number of triangles has been reduced from 103 to 54.
- Booleaned subbranches into main branches, with adjustments so they weren't visibly clipping through.
- Same trunk unwrapping, leaf fixes & vcol removal, and deduplication work as on _01.
- Normals work, using G7's advice on leaves.
- Added LOD2 and replaced LOD1 with more severe decimation.
- Vanilla was 67313 bytes, previous MOP version 92980; now 45294. (33% reduction from vanilla, , 51% from previous MOP version.)

flora_tree_03
- Same work as on _04. Why 88 collision triangles when 56 do?
- LOD0 376 tris, LOD1 177.
- Vanilla was 23955 bytes, previous MOP 23171 bytes. Now 24348 bytes. (2% increase over vanilla, 5% increase over previous MOP version)

flora_tree_ai_05:
- Added missing branches to collision mesh, simplified, and resplit. Was 142 tris in 5 sections, now 163 in 7 sections.
- Tweaked branches (to avoid clipping and subtly fix unsubtle unrealism), then union-booleaned them with the trunk.
- Copied over pointed branch tips from LOD1 to LOD0 to fix completely open ended vanilla mesh.
- Leaf work, including catching a few which needed the backside texture mirrored. On two, the entire mesh needed turned around, as simply rotating the UVs would leaf the leave... I mean, leave the leaf group the wrong shape for the texture to fit at the same size.
- Normals work, using G7's advice on leaves.
- LOD0 964 tris, LOD1 669, LOD2 298
- Vanilla was 88305 bytes, previous MOP 113080; now 64130. (27% reduction from vanilla; 43% reduction from previous MOP)

flora_tree_ai_06:
- Simplified collision mesh. Corrected apparently accidental inclusion of a couple faces when splitting collision mesh, resulting in nearly as large a calculation volume as an unsplit mesh, in the process. Added simplified versions of side branches for verisimilitude. -- Was 88 tris, now 82.
- Exact same work as on _03 and _04.
- LOD0 656 tris, LOD1 496, LOD2 240
- Vanilla was 35961 bytes, previous MOP 31901; now 50979. (42% increase over vanilla, 60% over previous MOP)

Note: some portions of the leaves intersect with the trunk meshes. This probably costs performance, but I worried that simply moving them would change the look too much, and didn't want the tedium of doing so then shifting the UVs to try to meet the tree properly.