[core] Bump standard library to libstdc++5-dev and clang version to 3.8.

[ChrisLoer]

No description provided.

[mention-bot]

@ChrisLoer, thanks for your PR! By analyzing the history of the files in this pull request, we identified @brunoabinader, @kkaefer and @jfirebaugh to be potential reviewers.

[jfirebaugh]

This proved to be a prerequisite for portions of #6984, namely using <codecvt> for UTF transcoding.

@kkaefer @brunoabinader @springmeyer See any downsides to upgrading? This basically brings mapbox-gl-native in line with the recommendations of https://github.com/mapbox/cpp.

[springmeyer]

👍 This feels like the right way forward. #309 tracks when we hit this before and decided to work around without upgrading libstdc++. I think it is a good time to require at least libstdc++-5-dev.

See any downsides to upgrading?

I think the upsides outweight the downsides. And the downside is not a blocker, but I'll detail it for the record. It is that requiring libstdc++-5 instead of 4.9 will trigger a GLIBCXX symbol dependency will jump to GLIBCXX_3.4.21. This means that binaries built of mbgl-native will work out of the box on systems >= ubuntu xenial, but will not work on ubuntu trusty or ubuntu precise without upgrading libstdc++. Because we were targeting libstdc++-4.9 before, mbgl binaries (including node-mbgl) would run on ubuntu trusty by default. However, for Mapbox this has no impact, since we've already upgraded our systems to support GLIBCXX_3.4.21 as per the upgrade methods described in https://github.com/mapbox/cpp.

[kkaefer]

As an alternative to using codecvt, we could also make this a system-specific API and backfill with Nunicode on platforms where no platform-native alternative is available.

[jfirebaugh]

codecvt is a standard C++11 feature -- it makes more sense to me to use it rather than platform-specific equivalents.