Fix documentation on the nature of the vision system orientation data

[PeterJCLaw]

Note: what's described here relates to the orientation data and not to spatial coordinates (see #434 for those). This only considers april_vision in "legacy mode" as its "new mode" uses a wholly different coordinate space.

april_vision appears to match the behaviour of Zoloto as I documented the latter in RealOrangeOne/zoloto#312. However this does not match the system outlined in the SR docs.

Zoloto/April:

• rot_x is roll
• rot_y is pitch
• rot_z is yaw

Docs:

• rot_x is pitch
• rot_y is yaw
• rot_z is roll

I haven't yet compared the zero positions and directions, however given the discrepancy in axes it is likely that those are at wrong too.

[PeterJCLaw]

WillB97/april_vision#7 also complicates things somewhat

[PeterJCLaw]

#437 fixed this.