Modernize project + subset without pyftsubset

[valadaptive]

I've made a webapp for subsetting and converting fonts, and part of that included compiling HarfBuzz' hb-subset API (and woff+woff2 conversion) to WebAssembly and exposing it via JS. As a proof-of-concept, I decided to try porting glyphhanger over to it, and it works!

Part of that work involved porting the codebase to use ES Modules, and I've also replaced a lot of dependencies with lighter alternatives (get-stdin -> Node builtins, chalk -> picocolors, glob -> tinyglobby). I'd also like to replace serve-static with sirv, but it doesn't yet function the same.

There are some other changes:

• As part of getting the glyphhanger client script to work properly across Node versions, and because the new version of characterset is ESM-only, I've decided to just bundle the client script using esbuild into generated/glyphhanger-script.umd.js.
• The output font files in the tests seem to be a couple dozen bytes larger, possibly because hb-subset supports more tables instead of dropping them?
• I've removed the woff-zopfli format entirely; all WOFF1 encoding now always uses zopfli. I've also changed the default settings to no longer emit WOFF1, since all modern browsers support WOFF2.

The documentation could probably stand to be updated, and I'm waiting for that sirv PR to get merged so I can switch this project over to it. But I'm interested in feedback.