Add in-situ cartesian field map construction from Geant4

[esseivaju]

Implements CartMapMagneticField as a G4MagneticField wrapper around Celeritas CartMapField. Includes:

• CartMapMagneticField class using pimpl to hide C++20 dependencies
• MakeCartMapFieldInput function to query and build Celeritas field from an existing G4Field
• Refactored CylMapMagneticField to use shared field sampling utilities
• Added covfie dependency to accel

[github-actions]

Test summary

 4 456 files   6 838 suites   13m 34s ⏱️
 1 850 tests  1 841 ✅  9 💤 0 ❌
23 540 runs  23 455 ✅ 85 💤 0 ❌

Results for commit 1014963.

♻️ This comment has been updated with latest results.

[sethrj]

Looks great! This is a nice addition. @rahmans1 @sanmayphy this should be useful in your framework integration.

[sethrj]

Thanks!

[sethrj]

I did not do a thorough job with this review. I know we were in a rush to get this in, but this code is virtually untested. I can't edit any of it for fear of breaking something. Our helper functions are using different units internally. ☹️

[esseivaju]

Did you spot a bug where we have a unit mismatch? I can take a look at adding unit test but since we're getting the field from Geant4, some part would require a mock Geant4 run

[sethrj]

While working on #2209 I needed to touch convert_to_geant and convert_from_geant. This led me down a rabbit hole into the field code since I couldn't test it at all, and I think I see an order-of-10 magnitude error when we use CGS but not when we use CLHEP...

Never mind, I am an idiot and it was a problem with my test: I had the wrong units for the scale factor (neglecting a 1/length).