[CURA-10464] Tree Support 2.0 speed-ups

[rburema]

This is mostly just two parts:

• Upping (and fixing) the minimum step size in the safeOffset___ functions, since that does sort-of the same as a normal offset, except then in steps, and each step also has a union or difference with the collison. The less you do this the better for speed, but you can't also just make it a normal offset, since that could make the support polygons 'move' through the model.
• Decreasing the amount of work done in order to get all of the circle polygons merged into one big blob. This is done by a) reducing the resolution of the individual 'circles' (now dodecagons I suppose), and b) not unioning the circles all at the same time, but instead in iteratively split groups.

I wanted to make the suggested benchmarks as well. However, since most tree-support classes worth benchmarking take the entire slice-storage and we agreed on a time-window for this ticket, I'll park that for now.

[github-actions[bot]]

Unit Test Results

25 tests  ±0   25 :heavy_check_mark: ±0   15s :stopwatch: -1s   1 suites ±0     0 :zzz: ±0    1 files   ±0     0 :x: ±0 

Results for commit 19de6e8c. ± Comparison against base commit d8e69924.

:recycle: This comment has been updated with latest results.