Support for separating TTC font when serializing

[harry75369]

Current implementation of embedding currently used typefaces into VGG file depends on SkTypeface::serialize which will save entire TTC data, rather than the subset.

(Some prerequisites: A TTC file is a collection of TTF fonts. Each TTF font maps to only one skia typeface.)

In order to minimize the saved VGG file, we need to subset the TTC data for target typeface. This could be achieved in two different ways:

1. when loading TTC file, separate it into TTF files and load each typeface accordingly. Or
2. when serializing certain typeface, extract the correct part of original TTC data which is stored in SkTypeface and then make the serialization.

The first way is preferred for implementation, but it will increase memory usage since TTC data is usually shared between all typefaces.

Some implementation references:

• https://github.com/yhchen/ttc2ttf
• https://github.com/oysterlab/ttc2ttf