Extend BIH capabilities for intersection operations

[elliottbiondo]

This MR creates the BIHIntersectingVolFinder class which is used to find the surface that a ray intersects with first, and the corresponding distance to this surface. This is done using a traversal strategy that is distinct from BIHEnclosingVolFinder, as is documented in their respective top-level doc strings. In a subsequent MR, this will be "hooked up" for use in background volumes.

[github-actions]

Test summary

 3 901 files   6 012 suites   4m 13s ⏱️
 1 617 tests  1 588 ✅ 29 💤 0 ❌
20 000 runs  19 925 ✅ 75 💤 0 ❌

Results for commit acc261d.

♻️ This comment has been updated with latest results.

[elliottbiondo]

@sethrj this is ready. Probably better to review this one first, then I will do the vol_id + others one.

[sethrj]

Still working on this, but gotta take the day off and here's what I have so far.

[sethrj]

One final comment is to consider whether it would be better to split this class into class BihIntersectFinder and template<class F> class BihIntersectingVolFinder, which I think would have a more sensible interface. (If we do that we should also add a note in BIHEnclosingVolFinder that we could update its interface in the future since I know that's what we based this on.)

[elliottbiondo]

@sethrj this is ready for another look.

I don't understand what you mean by splitting this into class BihIntersectFinder and template<class F> class BihIntersectingVolFinder. I don't think there will ever be any other form of BihIntersectFinder other than the volume version.

[sethrj]

I don't understand what you mean by splitting this into class BihIntersectFinder and template<class F> class BihIntersectingVolFinder. I don't think there will ever be any other form of BihIntersectFinder other than the volume version.

I just mean this as an implementation detail: half the functions are templated on F (which is probably better as a class template parameter) and the other half aren't (so they probably shouldn't be part of a templated class). One solution would be to move the non-templated to an "implementation base class" and then template the class on F.

[sethrj]

Looks good after you add the last min_dist and change double to real_type. Thanks!