Replace monaco-editor

[texodus]

A new implementation of <perspective-viewer>'s Expression Editor component, which replaces monaco-editor. While monaco-editor has lots of enviable features, it also had some issues:

• monaco-editor packages and imports it's own CSS, which causes issues when building with CSS-aware tools like webpack and required special consideration during installation, in general making the first-impression experience poor.
• While lazy-loaded (not downloaded until the user clicked the New Column button), monaco-editor was 2.6mb, by far the largest component in the project, and parsed/loaded much slower than similarly-sized WASM assets as it is JavaScript. The non-lazy part still imparted a ~100k JS payload in the main bundle.
• While the features we used were great, many features provided by monaco-editor are just not needed for the much smalled scope of an expression editor (e.g. "Rename Symbol" is probably not worth the code download and I doubt many users even knew it was there, hidden in the right-click or command-P menu).
• It had an awkward API due to the Worker and language support, requiring importing some strangely structured modules and unclear-if-global options. The examples are complex and show a lot of diverse use cases, but the actual documentation beyond that is just the raw API docs themselves which are quite sparse. This all required a lot of gnarly Rust FFI code to integrate.

The new version has been implemented-from-scratch:

• Weighs in at 35k in total (including the new Rust libraries added like nom). When accounting for the monaco-editor Rust FFI code removed, even the WASM asset is actually slightly smaller than in 1.6.7:

• Requires no special build accommodations at all.
• Features:
  • Syntax Highlighting
  • Error indicator (though the UX has changed since the monaco-editor version as the message is now displayed in the status bar, rather than in a tooltip).
  • Autocomplete dropdown for functions (and added their descriptions inline).
  • Line numbers

There are still some missing features.. Given the developer-experience impact of this change, I think its worth merging even as-is, with an eye towards addressing these shortcomings in the next few releases:

• No column name auto-complete
• No ident-on-new-line
• No paren-matching
• Autocomplete doesn't handle keye vents correctly yet