Fix T85948 Exact boolean crash with some nonplanar ngons.

Triangulating ngons could fail with the method that was being
used: projecting along the dominant normal axis and then using CDT.
It could fail if the ngon has self crossings or might be so after
the described projection.
Switched to using projection along the normal itself, and also to
using polyfill which produces some kind of triangulation no matter
what in such circumstances. This will also likely be faster if
there are a lot of ngons in the meshes, since the exact arithmetic
CDT was being used before, and now float arithmetic is used.

Get branch into compling state again.

There are some changes in Materials in the exporter that make compiling
the importer fail. I fixed one and several others are just ifdef'd out
for now so that the branch compiles. TODO to fix the rest.

Merge branch 'master' into soc-2020-io-performance

Fix compilation error in bypass bmesh commit when GMP not defined.

Added a cast needed to shut up an error in last commit.

Change Exact Boolean modifier to skip round trip through BMesh.

The Exact modifier code had been written to avoid using BMesh but
in the initial release the modifier still converted all Meshes to
BMeshes, and then after running the boolean code on the BMeshes,
converted the result back to a Mesh.
This change skips that. Most of the work here is in getting the
Custom Data layers right. The approach taken is to merge default
layers from all operand meshes into the final result, and then
use the original verts, edges, polys, and loops to copy or interpolate
the appropriate custom data layers from all operands into the result.

Forgot to free a bvhtree with previous commit.

Fix T84493 et al: New Boolean on Suzanne.

While Boolean is not guaranteed to work if the operands are not
volume-enclosing (technically: PWN - piecewise constant winding number),
it needs to do something in those cases. This change makes
more cases meet user expectations in T84493, T64544, T83403,
T82642 (though very slow on that one).
The original new boolean code used "generalized winding number"
for this fallback; replaced this with code that uses raycasting.
Raycasting would have been faster, but for unfortunately also
switchd to per-triangle tests rather than per-patch tests since
it is possible (e.g., with Suzanne) to have patches that are
both inside and outside the other shape. That can make it much
slower in some cases, sadly.

Fix T84465 Bevel Percent and Absolute when limit type is Weight.

The previous fix to the width modes Percent and Absolute did
not take into account that with limit mode Weight, the amount
needs to be scaled by the bevel weight of the beveled edge in
question. (Sometimes there are two beveled edges in question,
in which case an average is used.)

Fix T84465 Bevel Percent and Absolute when limit type is Weight.

The previous fix to the width modes Percent and Absolute did
not take into account that with limit mode Weight, the amount
needs to be scaled by the bevel weight of the beveled edge in
question. (Sometimes there are two beveled edges in question,
in which case an average is used.)