Electron-main with ESM-only deps

[komret]

Configure Electron build process to output main and preload bundles as ES instead of CommonJS.
That is possible since electron 28.0.0.

It is needed to upgrade dependencies which do not support CommonJS anymore:

• electron-store@10.0.0 (this is for sure)
• long really? To be confirmed
• @noble/hashes@2.0.0 really? To be confirmed

Because we use bundler (webpack) this solves most of our problems, because the bundler takes ESM package and squashes it into output js file. So actually only libraries which cannot be bundled (e.g. have a .node binary) and need to have a runtime require are a problem.

[Lemonexe]

Status report: TL;DR: tried many things, invested a lot of effort, but did not come up with a viable solution yet ❌

Electron-store + CJS

Original problem: it's incompatible, as stated. Build runs fine, but in runtime, the electron-main script throws ❌

Webpack + ESM

WIP in #15544
Does not work with chunking, see the PR. Though it does successfully run without chunking ✅
Since chunking was the very reason for switching ESBuild to Webpack (main+preload) in #9931, it's not something we can write off easily.
➡ I'd like to make minimal reproduction example and report bug at Webpack. Maybe it can be fixed there..

Rspack exploration work

An emerging bundler that seeks to reimplement webpack in rust, being mostly compatible. They claim that projects with simple config can replace Webpack with Rspack 1:1 with almost no effort.

💭 🤔 Maybe, just maybe, ESM could work better in Rspack? Also, it could significantly improve DX - much faster initial dev bundle + HMR 😍
➡️ So I explored.

But in Trezor Suite, our webpack config is very complex. It required a considerable effort. I've made it work for main+preload with CJS, but I have not finished renderer.. See #16017
Anyway, since main+CJS works I tried ESM, but it's even more broken than in webpack (branch here) 🙈

The peformance benefits do look promising, about 50% quicker for dev builds.

Conclusion: I couldn't make Rspack work, and see no clear path forward. It was a lot of effort, but personally I'd write it off as sunken costs 😢