Corrects beamed cross-staff stems' pure heights

Corrects beamed cross-staff stems' pure heights

An incorrect subtraction made it such that the pure heighs were
too long for stems on extremal staves and too short on interior
staves.

[Keith, 2012-09-22 05:56:19]

LGTM of course

http://codereview.appspot.com/6543046/diff/1/lily/stem.cc
File lily/stem.cc (right):

http://codereview.appspot.com/6543046/diff/1/lily/stem.cc#newcode342
lily/stem.cc:342: // find the offset of various staves as well as the minimum
and maximum offset
How does this "find the offset of various staves"?   
Maybe the common_refpoint is the System, in the cases where the current Stem is
attached to a cross-staff Beam,
but then the code is still finding offsets of Stems, note Staves.

http://codereview.appspot.com/6543046/diff/1/lily/stem.cc#newcode351
lily/stem.cc:351: // group where the beam lies
Now you're writing too much comment.  Most people think it is a bad idea to
write what the code does, only the reason why.
// Estimate the Stem to extend to the level of the furthest
// note-head in within its beamed group.  If that head is in
// a different VerticalAxisGroup, the height assumes the 
// staff-spacing tentatively assumed before line-breaking.

[MikeSol, 2012-09-22 08:36:52]

I'm gonna hold off on posting to this one as it evolves constantly with my work
on issue 2801, which is revealing several issues in cross-staff stem
calculations.  Irrespective of how issue 2801 pans out, I'll take all the
relevant code from stem.cc and either push it as a separate commit or include it
in this Rietveld if my patch for 2801 isn't LGTM'd.