Refactor build_acceleration_structures to require fewer passes

[Vecvec]

Connections

Description
This PR take advantage of changes from encode on finish to reduce the number of temporary vectors and do BLAS processing in one less pass. Also swapped some Vec::news to Vec::with_capacity because the capacity is now known.

A notable benefit of this is that implementing custom geometry (AABBs) should be easier because previously the intermediate vector storage assumed that all geometries were triangle geometries.

Note: first commit & most of the second commit are refactors without changing names to try and make things easier to compare - I recommend reviewing the first one w/o whitespace diffs to check no validation has been removed. (These first few commits also don't compile) The next few commits are for changing the references.

Testing
Uses existing tests

Squash or Rebase?
Squash

Checklist

• Run cargo fmt.
• Run cargo clippy --tests.
• Run cargo xtask test to run tests.
• If this contains user-facing changes, add a CHANGELOG.md entry. No user facing changes.

[inner-daemons]

I'd like to look at this but almost more out of curiosity than because I think I can provide a proper review. So this shouldn't block a review by @cwfitzgerald.