Autobeaming patch rev 2

[Carl D. Sorensen]

I've revisited my proposed patch on autobeaming.

The time signature for the autobeaming is now taken from
timeSignatureFraction instead of from measureLength and beatLength.  This
prevents the previous conversion of 12/16 to 6/8.

In order to deal with potential conflicts between autobeaming based on
beatLength and beatGrouping, I have established the following logic.

1) if beatGrouping is valid (meaning that the total number of beats in
beatGrouping times beatLength is equal to measureLength), then beatGrouping
is used for autobeaming.

2) if beatGrouping is invalid, then beatLength is used for autobeaming.

The effect of this is to keep all existing code that controls autobeaming by
beatLength working, because when beatLength is changed, beatGrouping becomes
invalid.

Obtaining the time signature from timeSignatureFraction resolves issue 511.

I've attached a file that shows how this works (it could become a regression
test), a diff of scm/auto-beams.scm, and a pdf of the output.

Is it OK to apply?

Carl

auto-beam-2.patch
Description: auto-beam-2.patch

beatGrouping-test.ly
Description: beatGrouping-test.ly

beatGrouping-test.pdf
Description: beatGrouping-test.pdf