The hard coded value was too small for certain calculations. Better to
have a configurable value that is initially set to our error level to
allow for deviations that don't meet the visibility test for spikes.
These have become more apparent with Clipper2
1) Move a bunch of std::map's to std::unordered_map to get constant-time
look-ups
2) Lengthen progress-reporting intervals to spend more time doing work
and less time talking about it
3) Reverse order of SHAPE_LINE_CHAINs in thermal intersection checks to
make (much) better use of bbox caches
4) Don't re-generate bboxes we already have
5) Fix some autos that weren't by reference (and were therefore copying
large datasets)
6) Rename delta progressDelta so it's easier to search for in future
7) Get rid of a few more autos (because I don't like them)
8) Pass large items to lambdas by reference
Fixes https://gitlab.com/kicad/code/kicad/issues/12130
Thread pools are long-lasting executors that have close to zero overhead
when launching new jobs. This is advantageous over creating new threads
as we can use this for threading smalling jobs and smaller quanta. It
also avoids the heuristics needed to determine the optimal number of
threads to spawn
Very small vertices (length <= 2 or 3 iu) are frequent in filled areas, and
create false positive detections because one cannot calculate a meaningful
orientation of these vertices.
Seth Hillbrand
Jeff Young
Marek Roszko
Wayne Stambaugh
jean-pierre charras