Part of https://github.com/harfbuzz/harfbuzz/issues/1017

Part of https://github.com/harfbuzz/harfbuzz/issues/1017

For consistency.

Part of https://github.com/harfbuzz/harfbuzz/issues/1017

Fixes #779 and follow up to b01d9b3d

https://bugzilla.gnome.org/show_bug.cgi?id=659212 for more context

To revive Travis CI bot

It was ignoring the restriction that gids must be consecutive to be placed into a single group.

As Khaled's finding on #981 this is the way Uniscribe detects those fonts. This is its detection part.

Splitted from #950 and #986 IIRC Chromium had a policy about this encouraging it,
not sure about automated way to detect and add them but for now lets have the needed
ones of them.

It uses floor implicitly without it but explicit use of round will be more correct.

[subset] sketch out support for a call that lists all codepoints present in a font. Implement support for it in format 4 cmap sub table.

Obviously one can use a newer NDK for building an updated HarfBuzz
instead that now pretty old version however I am concerned if that
version checking is working with clang that is used with the newer
NDK versions.