Fix T86768, bevel doesn't loop slide sometimes.

Six years ago, Bug T44961 about unwanted spikes had me not do a loop
slide if the angle was too extreme, to avoid unwanted spikes.
The current bug showed that that angle was much too big, and limited
desired behavior in many cases. Changing the angle from 0.25 radians
to 0.0001 radians (about 0.006 degrees) still fixes the original bug
and seems very unlikely to be limiting desired behavior now.

Fix T89391, etc. Boolean bugs when objects have negative scale.

The old modifier code, now just used for Fast, has code in it to
flip faces of arguments when their tranform's negativity differs
from the main object's transform's negativity.
I had neglected to put that logic in when I made the change that
skipped the round trip through BMesh.
Fixing this means that the results are more what the user expects
when some or all operands have negative scales.

Another slight increase in speed for Delaunay CDT.

When the new "need_ids" flag is false and the output type is not
one of the valid BMesh kinds, there is no need to propagate even
a dummy id to all of the faces.

Fix bug in assert in delaunay test.

Assert was trying to say x coords of arcs lined up, and didn't do that.

Fix crash in delaunay C interface test.

The test forgot to set the new need_ids field, which luckily
exposed a bug in the C api for delaunay when that field is false.
Fixed the bug and the test, and added a test for the need_ids false
case.

Update documentation string for mathutils.geometry.delaunay_2d_cdt.

Make it optional to track input->output mapping in delaunay_2d_calc.

Some uses of delaunay_2d_calc don't need to know the original verts,
edges, and faces that correspond to output elements.
This change adds a "need_ids" value to the CDT input spec, default true,
which tracks the input ids only when true.
The python api mathutils.geometry.delaunay_2d_cdt gets an optional
final bool argument that is the value of need_ids. If the argument
is not supplied, it is true by default, so this won't break old uses
of the API.

On a sample text test, not tracking ids save about 30% of the runtime.
For most inputs the difference will not be so dramatic: it only really
kicks in if there are a lot of holes.

Greatly improve speed of Delaunay when have a lot of holes.

Using part of a patch from Erik Abrahamsson, this replaces the
use of linked lists for original id tracking by Sets.
I had thought that the lists were unlikely to grow to more than
a few elements, but when the mesh has a lot of holes (whose
original ids go *outside* the hole, and therefore, most of the
mesh), this assumption can be very wrong.
On a Text regression test, the time went from 11.67s to 0.16s
with this fix. I also tested to make sure that Boolean didn't
slow down with this, and found it actually had a very slight speedup.

Using Sets exposed a dependency on the ordering of the items
in the id lists, luckily caught by a mesh intersect regression test,
so fixed that.

Fix a compiler warning on Windows for Exact Boolean.

For some reason, the Windows compiler didn't like the
static function being used in the parallel_reduece.

Fix a compiler warning on Windows for Exact Boolean.

For some reason, the Windows compiler didn't like the
static function being used in the parallel_reduece.