Excludes grobs supported by cross-staff grobs from pure-relevant calculations

Excludes grobs supported by cross-staff grobs from pure-relevant calculations

[MikeSol, 2013-08-13 17:58:52]

Allows all items to be pure relevant

[Keith, 2013-08-14 05:34:18]

https://codereview.appspot.com/12857043/diff/3001/lily/axis-group-interface.cc
File lily/axis-group-interface.cc (right):

https://codereview.appspot.com/12857043/diff/3001/lily/axis-group-interface.c...
lily/axis-group-interface.cc:504: && dynamic_cast<Spanner *> (elts[i])))
So if a spanner lies over just one cross-staff item, we ignore the entire
spanner for purposes of estimation of the space needed by its staff ?  The
spanner still takes up space, we are just slightly uncertain of its position,
due to the possibility that upon final page-layout the position of the
cross-staff item might force the spanner to shift.  

It would seem better to deal with the concept of supported_by_cross_staff_grobs
in side-support-interface, where we are finding, or estimating, the position of
the spanner so that it clears its 'support's.

https://codereview.appspot.com/12857043/diff/3001/lily/side-position-interfac...
File lily/side-position-interface.cc (right):

https://codereview.appspot.com/12857043/diff/3001/lily/side-position-interfac...
lily/side-position-interface.cc:283: && cross_staff
Under these conditions,
 trying to estimate before line-breaking  (pure)
 where an object 'me' will align vertically (Y_AXIS)
 against an object 'e' whose position depends on staff-spacing (cross-staff)
we might want to skip this object 'e', whether 'e' is a stem or not.

If it is a stem, since we know it is cross-staff, is the test on the next line
of the stem's direction going to be reliable ?  It seems best to skip the stem
and estimate the alignment against the remaining objects in 'support'.

[mike7, 2013-08-14 05:37:00]

https://codereview.appspot.com/12857043/diff/3001/lily/side-position-interfac...
File lily/side-position-interface.cc (right):

https://codereview.appspot.com/12857043/diff/3001/lily/side-position-interfac...
lily/side-position-interface.cc:283: && cross_staff
On 2013/08/14 05:34:18, Keith wrote:
> Under these conditions,
>  trying to estimate before line-breaking  (pure)
>  where an object 'me' will align vertically (Y_AXIS)
>  against an object 'e' whose position depends on staff-spacing (cross-staff)
> we might want to skip this object 'e', whether 'e' is a stem or not.
> 
> If it is a stem, since we know it is cross-staff, is the test on the next line
> of the stem's direction going to be reliable ?  It seems best to skip the stem
> and estimate the alignment against the remaining objects in 'support'.

Great ideas - I like this better.  Will do.