Quantize a number of measurements to 1/200in for better rasterization

Quantize a number of measurements to 1/200in for better rasterization

While 1/100in would be a nicer common denominator, that's 0.7227pt and
a basic thin line tends to be 0.4pt already.  1/200in at least gives
consistent staff lines at 600dpi (a common printing resolution) and at
even multiples of 100dpi (used for antialiasing when rendering to
100dpi images).

Rounding to a known set of sizes at the typesetting stage instead of
at rasterization has the advantage of producing consistent line
thicknesses without corner geometry mismatches, always presuming that
the final rasterization is compatible with the assumption.  This
should result in default staves that have uniform line appearance
after rasterization (different staves as opposed to different staff
lines of a single staff might still be placed a non-integral number of
raster lines apart, rasterizing differently).

The disadvantage is that the variation of thicknesses and features for
high quality photo typesetters will not make full utilization of the
device.  However, for most output devices in common use, anticipating
typical resolutions is going to lead to more consistent results.


Also contains

Reduce default PNG resolution from 101dpi to 100dpi for better rasterization

The original resolution of 101dpi was chosen to make a staff space
7 pixels large.  However, staff spaces are now rasterized based on a
1/200in grid, so generating bitmaps at 101dpi would defeat the purpose
of that rasterization.



This issue is complementary to issue 2658 in that it tries dealing
with rasterization issues.  However, its main advantage will be with
print and bitmap output.  Since PDF previewers tend not to work with
standard output resolutions, positive and negative effects on PDF
previewers should mostly cancel out.

[lemzwerg, 2013-09-06 09:28:38]

LGTM, save the 72pt vs. 72.27pt issue.