[GoodPie]

Fixes the ManifoldMesher access violation in #45 and the freezes/crashes from negative inputs in #41.

• [Bug]: ManifoldMesher.mesh_tree crashes with an access violation when a branch level's resolution is below 1 #45: get_origins counts a branch's first segment twice, so branches shorter than two segments got a negative segment → NaN directions in the gravity pass → out-of-bounds reads in the mesher. The segment is now clamped to zero length. The double-count fix itself would change every saved tree, so it's deferred to the [Feature]: Leaves and End Branch Radius #18 work.
• Mesher backstop: non-finite branch angles/twists raise RuntimeError instead of reading out of bounds, and child rings keep at least 4 vertices (a negative radius used to give 2).
• [Bug]: User input not sanitized properly, negative values cause Blender to crash #41: Trunk length/resolution, Growth iterations/apical dominance/lateral density and Pipe Radius end radius/constant growth raise ValueError for out-of-range values instead of looping forever or crashing.
  Testing:
• New tests/unit/test_crash_regressions.py: the [Bug]: ManifoldMesher.mesh_tree crashes with an access violation when a branch level's resolution is below 1 #45 script for seeds 1–12, plus the negative-input cases.
• Locally:
  • 101 trees (presets, defaults, growth, pipe radius) are byte-identical before/after.
  • Under ASan, [Bug]: ManifoldMesher.mesh_tree crashes with an access violation when a branch level's resolution is below 1 #45 goes from 11/12 heap overflows to 0.
  • ctest and the pytest unit tests pass.

Merge after the CI/packaging PR: lint needs PR 1's formatting fix and the wheels need cmake.verbose removed.

Fixes #45. Refs #41.