Port to N-API

[springmeyer]

A fun task would be to start a branch of skel that switches NAN for N-API.

This branch would:

• Drop support for node versions (in the travis.yml) that don't support N-API
• Drop usage of NAN

The goal of this branch would be:

• Help us learn about N-API
• Help provide simple examples of N-API usage for the time when we want to (potentially) move all modules to it.
• Learn what we don't know we need to learn about N-API (this is probably the most important one)

Caveat: The reality is that skel will need to support older node versions for a while, so this branch would need to stay in a branch and might drift/get out of date. And it would naturally have tons of conflicts with master.

[springmeyer]

This remains relevant and important to learn about. I'm adding this to my list of things to look into. My plan is to find 30 minutes to experiment. I will create a branch and try porting things over and see how far I can get. /cc @GretaCB

[springmeyer]

With the release of node 10 this week (https://nodejs.org/en/blog/release/v10.0.0/), N-API is now no longer experimental (mapbox/node-pre-gyp#345 was key to getting "out of experimental status") so this means:

• @mapbox/core-tech should plan to port all our modules to N-API
• the first step I think should be to port node-cpp-skel to N-API

[springmeyer]

done in #161 🎉