improved error handling

[aoeftiger]

cf. rdemaria#33

To be improved before merging:

1. integrate error handling for PySixTrack already into loader_madx functionality -- by reading out errors via new cpymad functionality, e.g.:

>>> seq = madx.sequence.XYZ
>>> multipole = seq.expanded_elements[0]
>>> multipole.align_errors
AlignError(dx=-2.1754812916408347e-05, dy=-6.556356908596416e-06, 
ds=0.0, dphi=0.0, dtheta=0.0, dpsi=0.0, mrex=0.0, mrey=0.0, mredx=0.0, 
mredy=0.0, arex=0.0, arey=0.0, mscalx=0.0, mscaly=0.0)

>>> multipole.field_errors.dkn
[0.0,
 1.1472076071601915e-06,
 5.002980147857331e-06,
 0.013090083041168831,
 -0.19713806015162913,
 -73.91912412030126,
 -9.796705181638819,
 0.0,
(...)
 0.0]

2. possibly remove current approach with error table read out (more cumbersome approach, bloated functionality in PySixTrack)

Originally posted by @aoeftiger in rdemaria#28 (comment)