Remove apparent SkIntToScalar(SkScalar) call in SkScalerContext_FreeType

Remove apparent SkIntToScalar(SkScalar) call in SkScalerContext_FreeType

This change seems to fix the following assertion failure we have been seeing
in the Skia_Linux_Fixed_Debug build since http://code.google.com/p/skia/source/detail?r=3059 :

drawing... bitmapfilters [540 330]
../include/core/SkFixed.h:67: failed assertion "n >= -32768 && n <= 32767"
Committed: https://code.google.com/p/skia/source/detail?r=3066

[epoger, 2012-01-18 18:39:34]

This change seems to alleviate the problem, but I'm not certain if we expect the
FT_BBox structure to contain SkScalars or always ints...

[Steve VanDeBogart, 2012-01-18 18:45:43]

https://codereview.appspot.com/5557049/diff/1/src/ports/SkFontHost_FreeType.cpp
File src/ports/SkFontHost_FreeType.cpp (right):

https://codereview.appspot.com/5557049/diff/1/src/ports/SkFontHost_FreeType.c...
src/ports/SkFontHost_FreeType.cpp:1425: x_height = bbox.yMax / 64;
I'd guess that the real problem is that you also need SkIntToScalar(64)

[epoger, 2012-01-18 18:53:56]

changed to use SkFDot6ToFixed() at Mike's suggestion

[reed1, 2012-01-18 19:14:57]

lgtm

[junov1, 2012-01-18 19:31:42]

Is it normal for the bbox to have such large values?  I think the failure we are
seeing here is a symptom of a deeper problem, which I will continue to
investigate. The canvas's transform matrix seems to be out of whack, and I think
something I put in SkDeferredCanvas might not be SkFixed friendly.