Increase Delaunator orient function robustness

[MFraters]

I found a corner case where in debug mode where the Delaunator orient function return a different result depending on whether it is in debug or release mode. This causes issues in the release mode, which is basically wrong. The issue comes down to that (qy - py) * (rx - qx) - (qx - px) * (ry - qy) is exactly zero in debug mode (so (qy - py) * (rx - qx) - (qx - px) * (ry - qy) < 0.0 is false) and the release mode is slightly less accurate and results in -3.48086e-21, which then returns true.

The solution I have gone for here is to check if the value is smaller that -epsilon. I guess it is not a perfect solution, but it seems to solve the problem.

For reference, these are the values of (qy - py) * (rx - qx) - (qx - px) * (ry - qy) and the results for debug mode left and release mode right.
(0.80218-0.793353) * (-1.78205 - -1.7908) - (-1.7908 - -1.79954) * (0.811008 - 0.80218) = 0 | (0.80218-0.793353) * (-1.78205 - -1.7908) - (-1.7908 - -1.79954) * (0.811008 - 0.80218) = -3.48086e-21

Playing with some of the flag could potentially help, but I just want it to be more robust, which this pull request achieves.

[github-actions]

Benchmark	Main	Feature	Difference (99.9% CI)
Slab interpolation simple none	1.433 ± 0.036 (s=320)	1.436 ± 0.037 (s=310)	-0.5% .. +0.9%
Slab interpolation curved simple none	1.201 ± 0.003 (s=370)	1.201 ± 0.003 (s=382)	-0.0% .. +0.1%
Spherical slab interpolation simple none	1.497 ± 0.004 (s=296)	1.496 ± 0.005 (s=308)	-0.1% .. +0.0%
Slab interpolation simple curved CMS	1.375 ± 0.002 (s=340)	1.375 ± 0.002 (s=317)	-0.1% .. +0.0%
Spherical slab interpolation simple CMS	2.362 ± 0.005 (s=180)	2.361 ± 0.004 (s=204)	-0.1% .. +0.0%
Spherical fault interpolation simple none	1.527 ± 0.005 (s=289)	1.526 ± 0.006 (s=303)	-0.1% .. +0.1%
Cartesian min max surface	3.423 ± 0.033 (s=138)	3.419 ± 0.021 (s=128)	-0.4% .. +0.2%
Spherical min max surface	9.672 ± 0.147 (s=50)	9.581 ± 0.132 (s=46)	-1.9% .. +0.1%

[codecov]

Codecov Report

Merging #479 (c3753b2) into main (d41db79) will not change coverage.
The diff coverage is n/a.

❗ Current head c3753b2 differs from pull request most recent head 6ce66d1. Consider uploading reports for the commit 6ce66d1 to get more accurate results

@@           Coverage Diff           @@
##             main     #479   +/-   ##
=======================================
  Coverage   90.85%   90.85%           
=======================================
  Files          91       91           
  Lines        6406     6406           
=======================================
  Hits         5820     5820           
  Misses        586      586           

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update d41db79...6ce66d1. Read the comment docs.