Investigate Javenode and/or JNpm

Right now, we integrate with node.js formatting tasks by finding a Node.js and NPM installation on the host machine

- using npm to install the required formatter, usually prettier, which is slow (#766)
- [JNPM](https://github.com/OrienteerBAP/JNPM) might be able to do this better
- starting a localhost server in java, and another in node.js, and passing strings back and forth between them
- [Javenode](https://github.com/caoccao/Javenode) would allow us to do this in-process.
- Upside: fancy [module intercepting](https://github.com/diffplug/spotless/issues/556#issuecomment-919613345) which could cooperate with something like JNPM
- Upside: passing strings within-process would be faster, but this is minor
- Upside: don't need to futz with Spotless finding a node.js installation
- Downside: can't futz with Spotless using a specific node.js installation which might match other non-java tooling

I don't plan to work on any of this, but I think #766 is a big problem. If anyone wants to experiment with the above, they are very welcome to! An important part for getting a PR merged will be a real-world project, where you can say "it was X seconds before, and it's Y seconds now". I don't want to swap things just to swap them.

If there's an easy way to keep the old mechanism and the new mechanism at the same time, that might be nice, but it's fine to remove the old mechanism as part of the PR too. Maybe the new mechanism dominates the old, and if it doesn't we can figure out how to get them to coexist after the fact.

#766 とリンクされている Javenode および JNPM プロジェクトを調査し、現在の Node.js formatter のワークフローと考えられる代替案を理解してください。完全な調査には、既存のメカニズムと提案されたメカニズムの実環境での比較を含め、変更前後の所要時間を報告し、古いアプローチと新しいアプローチを併存させるべきかどうかを説明する必要があります。

索引モデルが issue の本文から書いたものです。