============================
* The optimize spell now includes two new spells:
opt_collisiongeometry for optimizing triangle based collisions, and
opt_collisionbox for optimizing simple box collisions. This should
result in faster loads and probably also a small but noticable
performance improvement.
* Fixed opt_collisiongeometry for multimaterial mopps (reported by
wildcard_25, see issue 3058096).
* New SpellCleanFarNif spell (suggested by wildcard_25, see issue
3021629).
* Bad normals are now ignored when packing a bhkNiTriStripsShape
(fixes issue 3060025, reported by rlibiez).
* The opt_delunusedbones spell no longer removes bones if they have a
collision object (fixes issue 3064083, reported by wildcard_25).
* If the jobs option is not specified in the toaster, then the number
of processors is used---requires Python 2.6 or higher (suggested by
chaky2, see issue 3052715, implements issue 3065503).
* New opt_delzeroscale spell to delete branches with zero scale
(suggested by chaky2, see issue 3013004).
* The opt_mergeduplicates spell now ignores (non-special) material
names, so identical materials with different names will get merged
as well (suggested by chaky2, see issue 3013004).
* New spell to fix subshape counts (see issue 3060025, reported by
rlibiez), it is included in the optimize spell.
* New opt_collisionbox spell which automatically converts triangle
based box collisions to primitive box collisions, which are much
faster in-game (contributed by PacificMorrowind).
* Optimizer spell now uses triangles to represent skin partitions
(improves in-game fps).
* Better vertex map calculation when calculating skin partitions
(improves in-game fps).
* Optimizer now always triangulates (improves in-game fps).
Stripification will be readded later in a modularized version of the
optimizer spell, for those that want minimal file size rather than
maximal performance).
* Much faster implementation of vertex cache algorithm (now runs in
linear time instead of quadratic time).
* Check triangle count when converting to box shape (fixes issue
3091150).
* Bugfix in vertex map reordering (fixes most nifs reported in issue
3071616).
* Bugfix in vertex cache algorithm (fixes a nif reported in issue
3071616).
* Cover degenerate case in ATVR calculation when there are no vertices
(fixes a nif reported in issue 3071616).
* Cover degenerate case when calculating cache optimized vertex map
(fixes a nif reported in issue 3071616).
* Remove branches if they have no triangles (again fixes a nif
reported in issue 3071616).