Tyler/inf2

[tylerflex]

Before, we made td.inf = np.inf and it worked, except for in Simulation. _filter_structures_plane() where it has to check intersections for the polygons against each other.

To fix, I added a way to evaluate a shapely.geometry.Polygon containing "inf" vertices.

The function looks at each vertex of the polygon and creates a copy where any infinite vertices are replaced by a signed td.constants.LARGE_NUMBER.

Then this is inserted into the Simulation. _filter_structures_plane() to avoid any problems in the plotting.

Might need to check it works for Circles as well.

[tylerflex]

Alright, this should work but doesn't.

Seems there's some errors in the assert in discretize_ind.

But I can't reproduce, it only happens on the backend, it seems.

One notebook that triggers it is GratingCoupler.ipynb.

[tylerflex]

@momchil-flex this is ready for testing.

• Plotting now supports objects with td.inf size.
• PolySlab.slab_bounds is inf-aware.
• Box.from_bounds is inf-aware.
• center and radius are prohibited from being inf, to avoid inconsistent behavior.
  tests pass, but we should maybe run backend tests with this branch too, to catch runtime errors.