Split BSVertexData into two versions, one for FO4 and one for SSE.  This
uncomplicates some of the conditions and lowers the row count for each
vertex compound so will be faster to parse.

Also some misc name/type changes for FO4 blocks.

Renamed to NiBound to reflect the actual Gamebryo type.

This repeats the previous 32 materials but with `_PLATFORM` and
`_STAIRS` to reflect the bit 5 and 6 comments in the description.  This
makes them user selectable whereas before there was no way to set these
materials.

NiTimeController, BSBehaviorGraphExtraData update

Num UV Sets fix.  Merging as I have used this for months in NifSkope without issue.  Handling it as one ushort instead of two bytes is the only correct way due to endianness issues.  Addresses all regressions with Bethesda NIFs as well as maintaining non-Bethesda NIF support.  Closes #51 

This is a redo of #51 since I did not seem to have permission to rebase the PR.   These changes work for all games and fix all FO3 UV Sets regressions.

The only downside is for UV Sets > 1 in non-Bethesda games the UV Sets count now becomes a combination of flags.  UV_1 + UV_2 = 3 UV Sets,  and UV_1 + UV_4 = 5 UV Sets and so on.

This was the only way to treat the field as a single ushort which is necessary for endianness changes when reading the field 16 bytes at once because the value is read/written 16 bytes at once and uses bitmasking to retrieve the UV count.

Motor types decoding from https://github.com/jonwd7/0x6e6966746f6f6c73/blob/0532cd7bdffc10b578c8b8e841a0f6206c62d6ab/nif/bhk.bt#L367

Added motor descriptor to Prismatic constraint.

Added motor variable explanations, added default values and removed the white spaces.

Added motor type selector for Ragdoll and Limited Hinge constraint.

SSE Support

Support for SSE.  The BSVertexData had to be moved up for the
SkinPartition.  The changes have been tested on all SSE files and used
for the past several months with no issue.  There are no known FO4
regressions.

A new token for division in expressions was introduced, `#DIV#`, as "/"
is used in some XML attributes already and also causes parsing issues
with NifSkope which already uses that token for things at runtime.

The BSSegment PR from ttl269 has also been included in this as it is a
valid change and required for SSE.

Removal of nifskopetype

In <basic> and <compound> the nifskopetype attribute was used as an
alias system to remap types to internal types.   NifSkope now handles
the remapping internally.   One single downside to this is that new
basic/compound cannot be remapped to an internal type via XML alone, it
must be added to NifSkope.  However, if this proves to ever be an issue,
NifSkope can house another file in the installation where the remapping
can be done.

Additionally, nifskopetype was used in some <add> rows to denote that
the data was a binary blob.  I have instead replaced this with a more
generic `binary="1"`.  This is also more informative as generally these
rows are actually embedded files.

Lastly I removed a lot of custom "ns" compounds which NifSkope no longer
seems to use.  It would seem these compounds were even pre-nifskopetype
attribute.

FO4 Support

There were 16 changed shader flags between game versions.  This creates
a new set of enums for FO4 so that the proper descriptions can be shown
for the bits.

Also, the original vercond for Shader Flags 1/2 was unnecessary.  The
Skyrim NIFs that are User Version == 11 are actually structured like FO3
NIFs and use different shader blocks.  So changing the condition like I
did is a sound change, and I've already scanned all vanilla Skyrim NIFs.

Also changed inheritance for bhkNPCollisionObject because the Flags are
different now.  Also bhk*System aren't actually extra data so I changed
the inheritance there.

Names were again changed to mimic what they are called in the engine.
Some previously unknowns were clarified as well.

Changed to match the names found in the engine.

After changes to NifSkope parsing there is no longer a huge benefit to
bypassing all the cond="" / ARG checking on the BSVertexData.  The other
programs did not have this issue, so the variants can be removed.

Combined VF6 and VF7, created <bitflags> after rigorous testing.
Rigorous catch-all compound instead of the mess before.  Removed edge
case compounds as they are no longer needed.   Left optimized compounds
for NifSkope speed issues.

PyFFI does not like missing template attributes for Ref/Ptr despite the
documentation saying the attribute is optional.

Validation for "Skin" on BSTriShape is set only to NiObject for the time
being.  Setting it to BSSkin::Instance would require moving around the
blocks.

Correct data length in all cases, but not decoded.

Supports several edge cases at the cost of being more complicated since
the XML is not set up to allow testing of multiple flags in compounds.

Now loads SCOL with crashing NifSkope and without any errors.  The
PreCombined folder now at least opens but with errors as
BSPackedCombinedSharedGeomDataExtra is not decoded.

Much wider support of meshes overall, without any NaN issues stemming
from misassignment of half float types.

Very little actually changed, just the addition of 3 textures + env map
scale.