[brlcad-devel] when bad conversion stats get worse

So over the past month (32.87 days to be more precise), I've had the  
conversion script churning through a couple hundred real target  
models to see how bad things get.  They're considerably worse than  
the sample geometry (~2%) and preliminary testing (< 7%).  We're seem  
to be under 15% failure:

    Files:  190
  Objects:  604069
Failures:  95595 NMG, 99486 BoT
NMG conversion:  84.2%  (508474 of 604069 objects)
BoT conversion:  83.5%  (504583 of 604069 objects)
   Success rate:  83.9%

That implies real world models have 2x-3x the number of failures than  
we see with sample geometry. That just further reinforces the need  
for a complete revamping of the conversion and boolean evaluation  
process.  Also means that even if we get 100% conversion on sample  
geometry, that is only a start (albeit a good start).  At best, real  
world models just have more of the types of geometry that are failing  
due to their increased complexity.

Going to do another run of just primitives next since they're the  
lowest level failure that will cascade all the way up the hierarchy  
(and should be some of the easiest to fix).

Cheers!
Sean