experiment: collapse PosRot cfg within the struct

[danieleades]

the cfg flags on the PosRot struct can be moved onto specific methods of the implementation block, instead of duplicating the entire block between the glam and not(glam) features

leaving as a draft, for discussion

[VirxEC]

I did [f32; 3] as it guarantees that PosRot is 12 bytes. If this causes no perf regressions and makes implementing stuff like #6 easier, then I'd be down for this change.

[danieleades]

i guess it must still be 12 bytes or that test would be failing

[danieleades]

there's probably further optimisation that can be done here if including nalgebra, since PosRot is roughly equivalent to nalgebra's Isometry2