hide the Hinter implementation in the freetype/truetype package.

LGTM=bsiegert
R=bsiegert
CC=golang-codereviews
https://codereview.appspot.com/58940043

The logic is a little clunky, but as before, let's get it right first,
then get it clean once we have a full battery of tests.

R=bsiegert
CC=golang-codereviews
https://codereview.appspot.com/50910043

R=bsiegert
CC=golang-codereviews, remyoudompheng
https://codereview.appspot.com/39440045

The very subtle pp1x versus phantomPoints[0].X distinction is pretty
ugly code, but let's get all the tests passing first, so that we can
refactor with confidence.

R=bsiegert
CC=golang-dev, remyoudompheng
https://codereview.appspot.com/35850043

*before* hinting.

R=bsiegert
CC=golang-dev, remyoudompheng
https://codereview.appspot.com/35070044

R=bsiegert
CC=golang-dev, remyoudompheng
https://codereview.appspot.com/34700043

2nd / 4th phantom point rounding.

R=bsiegert
CC=golang-dev, remyoudompheng
https://codereview.appspot.com/32040043

Successfully hint all of DejaVu Sans Oblique, barring 4 exceptions
that are errors in the upstream C Freetype project. See
http://lists.nongnu.org/archive/html/freetype/2013-11/msg00004.html
for details.

R=bsiegert
CC=golang-dev, remyoudompheng
https://codereview.appspot.com/21580044

Also add API for vertical metrics.

R=bsiegert
CC=golang-dev, remyoudompheng
https://codereview.appspot.com/21330043

R=bsiegert
CC=golang-dev, remyoudompheng
https://codereview.appspot.com/19090043

Also make phantom point delta-X adjustment.

Also make print-glyph-points require Freetype version >= 2.5, which is
the latest version, at the time of writing.

R=bsiegert
CC=golang-dev, remyoudompheng
https://codereview.appspot.com/18370043

Also increase the maximum recursion depth from 4 to 8, since some
x-deja-vu-sans-oblique glyphs are actually depth 4. It's an arbitrary
limit. The C Freetype code doesn't seem to limit recursion (below
whatever the font says it needs), but it seems safer to have a limit.

R=bsiegert
CC=golang-dev
https://codereview.appspot.com/14494062

R=bsiegert
CC=golang-dev
https://codereview.appspot.com/14425064

one more test case.

R=bsiegert
CC=golang-dev
https://codereview.appspot.com/14691043

points accordingly.

Yes, it's a mess, but let's pass the tests first, and then refactor.

R=bsiegert
CC=golang-dev
https://codereview.appspot.com/14419052

Hinter's state, instead of having a glyph shrink and restore its
points around hinting.

R=bsiegert
CC=golang-dev
https://codereview.appspot.com/14203043

R=bsiegert
CC=golang-dev
https://codereview.appspot.com/14117043

R=bsiegert
CC=golang-dev
https://codereview.appspot.com/13855043

We can now hint the .notdef glyph from luxisr.ttf. Yay.

R=bsiegert
CC=golang-dev
https://codereview.appspot.com/12829048

Also add a test that freetype-go derives the same glyph points as
the C freetype implementation.

R=bsiegert
CC=golang-dev
https://codereview.appspot.com/12670046

These new opcodes aren't unit-tested per se, but they will be
exercised by an end-to-end hinting test of the Luxi fonts.

R=bsiegert
CC=golang-dev
https://codereview.appspot.com/12100043

R=bsiegert
CC=golang-dev
https://codereview.appspot.com/11983043

to package truetype.

R=bsiegert
CC=golang-dev
http://codereview.appspot.com/6427062

Thanks to Jeff R. Allen <jra@nella.org> for the conversation that
led to this change.

benchmark              old ns/op    new ns/op    delta
BenchmarkDrawString     21168440     20143860   -4.84%

The number of mallocs per iteration is unchanged.

R=rsc, r
CC=golang-dev, jra
http://codereview.appspot.com/6304077

No code changes, just a copy/paste.

R=bradfitz
CC=golang-dev
http://codereview.appspot.com/6201052