shortened flags: choosing appropriate flag (issue4410049)

[lemniskata . bernoullego]

Reviewers: ,

Message:
Passes regtests (i see a regression in partcombine-midi.ly, however from
Graham's message about midi regtests
http://lists.gnu.org/archive/html/lilypond-devel/2011-04/msg00283.html i
understand that it's not my fault).

If i understand correctly, the default way of calculating flags is using
c++ code which i modified, but there's also  another possibility - using
scheme. Looks like it's done in flag-styles.scm. However, i don't know
scheme; writing c++ code took me weeks (and a lot of help was required),
so i think implementing this algorithm in scheme is definately beyond my
skill.

Description:
choosing appropriate flag for shortened stems

Some stems are shorter than usual and require
special shortened flags.
This code searches through available flag variants
and chooses the best one.

extracting info about available flag glyphs

We need to know what length variants of flags
are available in the font used -
this might change with time and/or font used,
so it cannot be hardcoded.

Adding some shorter flags for testing

These flags are very rough and used only
to see if c++ code is indeed working.

Flag functions instead of defining glyphs directly

We will need many length variants of every flag.
Therefore instead of writing flag code directly
in glyph definition, it should be written as
a function and called back later as appropriate.
The argument shortening is the amount the flag
should be shorter than default.
As for now it is used in a very primitive way,
only to demonstrate the shortening effect
and to test c++ code on something.

Please review this at http://codereview.appspot.com/4410049/

Affected files:
  M lily/stem.cc
  M mf/feta-flags.mf