Spherical coordinates

see also: #543
@jenshnielsen
@MerlinSmiles
@WilliamHPNielsen

As discussed: My proposal for solving all spherical coordinate related issues is to write a single meta_instrument_driver which gets x, y and z parameters as input and handles the coordinate transformations internally but keeping the original phi, theta and r in its memory and setting the magnet x, y and z whenever phi, theta or r are changed.
support for tupled parameters would be a pre.
This would solve the problem for both the Oxford fridges and for the AMI magnets I think.

Review issue #543 and the existing instrument-driver approach before deciding on the design. The proposed driver should accept x, y, and z, retain phi, theta, and r, and update the magnet coordinates when those values change; determine whether tupled parameters can also be supported. Done means the coordinate behavior works for both Oxford fridges and AMI magnets.

Written by the indexing model from the issue text.