feat(babel): rewrite babel-preset-expo to use TypeScript

[EvanBacon]

Why

I found that if we increase the complexity and logic in our babel preset, then we can easily get to 30% faster web builds. Before doing this I'd like to ensure we have a better foundation to build against.

This PR converts babel-preset-expo to TypeScript and drops some safely unused features.

Plans

• We'll use the caller to determine if a module should be compiled for Node.js, if so, we'll use a much cheaper babel plugin.
• We'll add an additional caller parameter to detect if the module is inside the node_modules folder, if so, then we'll use a cheaper babel plugin. We'll also add the ability to allow certain node modules to be transpiled as application code.
  • The first version of this will be rolled out behind a flag to ensure we aren't breaking apps, since apps will more than likely break, as most React Native modules are shipped untranspiled.
  • Perhaps it makes sense to roll this out as a web-only feature.

Test Plan

Tests should keep passing.

[ide]

Worth making a plan to remove this over time? My understanding is that because developers can specify their own Babel configuration, we don't need to implement a flag to opt back in to adding this plugin as long as we explain the migration path.

[ide]

Out of scope of this PR but I think Brent mentioned this plugin causes warnings on installation, and there is a new plugin. Makes me think it is worth publishing a minor version upgrade of babel-preset-expo with these fixes, but would need to separate out the breaking changes around webpack.