Use matrices instead of rotors for transforms

[HactarCE]

reasons:

• in general, it's not actually easier to interpolate rotors compared to matrices
• matrices are better-studied so it's easier to find algorithms
• matrices scale better in high dimensions (rotors win in 6D and below; matrices win in 7D and above)

distance(A, B) = entrywise dot product

Notes (from things Milo was saying)

entrywise dot product of A and B = trace(A BH)

(BH = conjugate transpose of B)

for unitary matrices, BH = B^-1

trace = sum of diagonal = sum of eigenvalues

[HactarCE]

this leaves pga unused. not sure what to do with it