Fix memory leak in OpenGL visualizer

[jngrad]

Fixes #3530

Description of changes:

• free memory of quadric objects; they caused a 12 MiB/min memory leak in simulations containing bonds and a 3 MiB/min memory leak in simulations containing cylindrical shapes (Cylinder, SpheroCylinder, SimplePore, Slitpore)

• fix an issue in the visualization_constraints.py sample that prevented a Slitpore from being added to the list of constraints

[jngrad]

@christophlohrmann please check a few samples, e.g. the following are directly affected by this PR:

./pypresso ../samples/visualization_bonded.py --opengl
./pypresso ../samples/visualization_constraints.py --simplepore
./pypresso ../samples/visualization_constraints.py --slitpore
./pypresso ../samples/visualization_constraints.py --cylinder
./pypresso ../samples/visualization_constraints.py --spherocylinder

[codecov]

Codecov Report

Merging #3533 into python will decrease coverage by <1%.
The diff coverage is n/a.

@@          Coverage Diff           @@
##           python   #3533   +/-   ##
======================================
- Coverage      87%     87%   -1%     
======================================
  Files         536     536           
  Lines       24177   24177           
======================================
- Hits        21063   21057    -6     
- Misses       3114    3120    +6

Impacted Files	Coverage Δ
src/core/polymer.cpp	93% <0%> (-6%)	⬇️
src/core/electrostatics_magnetostatics/p3m.cpp	85% <0%> (-1%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update cd66da0...7747cab. Read the comment docs.

[christophlohrmann]

The shapes look like they are supposed to. Also memory doesn't blow up after longer runtime.
Still, in the samples, particles are on both sides of the constraints and also can hop across the constraints

[jngrad]

I've just tried to rewrite the particle placement loop to make some shapes impossible to cross, but it didn't work for the slitpore, where particles spawn too close to each other due to the confined geometry.

Particles crossing constraints is most likely a side-effect of the force capping mechanism, which prevents the force from becoming infinite when crossing a wall or surface. Writing custom particle placement code for each constraint and validating the simulation against crashes for each constraint in CI seems too much work for such a simple sample, and will make the sample a lot more difficult to re-use by novice users.

[KaiSzuttor]

you should use energy minimization instead of force capping

[jngrad]

We use steepest descent to minimize the energy, and then the Langevin thermostat with a high gamma. The force capping is there to prevent the simulation from crashing.