Avoids "deprecated" warnings.

Was introduced in glib 2.22.

Useful for profiling shapers.

As reported by Jonathan, OS X has setmode() that is something other
than what setmode() is on Win32.  So, limit invocation to Windows
platforms only.

Thanks to Khaled for spotting it.

Now, it looks like this breaks Cygwin again.  Have to figure out what's
going on.

Looks like Cygwin / MSVC declare it that way, and it still works on
MinGW32 cross.

Two output formats for now: TEXT, and JSON.  For example:

  hb-shape --output-format=json

Deserialization API is added, but not implemented yet.

Use with Arabic, for example, to see the effect on joining.

1. If there is any offset (x or y), print out both x and y offsets.

2. Always print out the advance in the major direction of the buffer.
Ie. even for zero-advance glyphs, print a "+0".  This is more intuitive.

Lohit Devanagari has a glyph named:
u0924_u094D.half_u0930_u094D.blwf.vatu

This reorders glyphs within the cluster to a nominal order.  This should
have no visible effect on the output, but helps with testing, for
getting the same hb-shape output for visually-equal glyphs for each
cluster.

Just turns all --show-* options on.

One less argument to type in typical testing workflow!

Also fix cairo cluster generation.

This means the --features indices also refer to char position
instead of byte position now.  Same for cluster values reported
by hb-shape.

Will add an option for byte indices later.

Sorry for the disruption but I need this to differentiate from the
Unicode string.

Ok, much more useful as a test suite driver now.

Patch from Khaled Hosny.