We need the font for glyph lookup during GSUB pauses in Indic shaper.
Could perhaps be avoided, but at this point, we don't mean to support
separate substitute()/position() entry points (anymore), so there is
no point in not providing the font to GSUB.

Comes with some 10% speedup for Devanagari even!

Fixes most failures of Oriya, and improves others a bit.

Experimental API for now.

Where causing issues for people with MSVC.

In old-style Indic OT standards, the post-base Halants are moved after
their base.  Emulate that by moving first post-base Halant to
post-last-consonant.

Brings test-shape-complex failures down from 88 to 54.  Getting there!

Fixes https://bugzilla.mozilla.org/show_bug.cgi?id=644184
among others.

Shapers now can request segmented feature application by calling
add_gsub_pause() or add_gpos_pause().  They can also provide a
callback to be called at the pause.  Currently the Arabic shaper
uses pauses to enforce certain feature application.  The Indic
shaper can use the same facility to pause and do reordering in the
callback.

Respectively, separate planner from the actual plan.

Previously boolean features turned on the entire feature mask.  This is
wrong if feature is Alternate and user has provided values bigger than one.
Though, I don't think other engines support such corner cases.

Previously if a default global feature was overrided by a non-global
user feature, we were not setting any default mask for the feature,
essentially disabling the feature by default.  Fix that.

hb_ot_plan_t may be a better name, donno.

We now build our entire attack plan before jumping in.