Fix Merge manifoldness issues

[elalish]

Following on from #474, I was still surprised how infrequently I could create a manifold from seemingly-solid input meshes. Turns out there were two major problems I've fixed here:

1. When vertices are merged together, this often results in two triangles being degenerate (not geometrically, but topologically - actually referencing the same vert twice). This was causing the mesh to be marked as non-manifold. The fix is to simply remove those triangles, which allows many input meshes to show up manifold that weren't before.
2. I finally found the problem with the Khronos-DragonAttenuation model. It was infinite-looping in the decimator. Turns out it was because of a pinched vertex (often called a non-manifold vertex - think two cones that share the same tip vert). I added a function to remove these, as they will also mess up the Genus calculation.

I also discovered how to debug Manifold right in Chrome, which actually works really well. I added instructions to the README.

[codecov]

Codecov Report

Patch coverage: 100.00% and project coverage change: +0.04 🎉

Comparison is base (4b65ff3) 90.27% compared to head (37cb2be) 90.31%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #481      +/-   ##
==========================================
+ Coverage   90.27%   90.31%   +0.04%     
==========================================
  Files          35       35              
  Lines        4318     4338      +20     
==========================================
+ Hits         3898     3918      +20     
  Misses        420      420              

Impacted Files	Coverage Δ
src/manifold/src/impl.h	72.72% <ø> (ø)
src/manifold/src/edge_op.cpp	98.04% <100.00%> (+0.09%)	⬆️
src/manifold/src/impl.cpp	95.42% <100.00%> (+0.05%)	⬆️
src/manifold/src/properties.cpp	84.81% <100.00%> (+0.09%)	⬆️

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.