Stem lengths on some beams seem too long

[ajyoon]

Beam lengths look good in the examples/beams.py example, but it seems the beams doc page examples run into a case where the beams are too long.

staff = Staff(ORIGIN, None, Mm(100))
Clef(ZERO, staff, 'treble')
chords = [
    Chordrest(Mm(0), staff, ["c"], (3, 16)),
    Chordrest(Mm(10), staff, ["e"], (1, 16)),
    Chordrest(Mm(20), staff, ["g"], (1, 8)),
    Chordrest(Mm(30), staff, ["bb"], (1, 8))
]
BeamGroup(chords)

[ajyoon]

On closer examination it seems that this is actually a fairly narrow edge-case that seems to happen when:

• Notes in a beam group slope in one direction over a fairly long interval
• The beam direction is the same as the note slope
• and it's made worse by having beamlets

This is caused by a constellation of factors in the beaming algorithm. It checks the closest beamlet's Y position - in this case the 16th hook on the E - and ensures the note closest to the beam (B-flat here) is some minimal distance away from it. Normally this produces decent results, but the above factors make the beam appear visually far from every individual notehead.

Given that this seems like a fairly narrow case and it may be an involved fix, I think we'll put this on ice for now and just modify the docs to show a more typical case. If you're a user who is annoyed by this bug please let us know and we'll try to prioritize it.