Recoil should rotate the camera, not adjust the forward vector directly

[iamgreaser]

As evidenced in the function at 0x00422850 in the original AoS, those tiny values are actually angles fed into an FSINCOS opcode.

In my own experiments, I've yet to make sense of the sqrt(1-y^4) thing though - when I try to make use of it, the recoil comes out wrong. Maybe I need to normalise the forward vector first...