Packaging enhancements

[SleepyLeslie]

• Adds Windows portable build - partly addresses Portable grist on Windows and simple configuration. #27.
• Packaging pipeline now produces artifacts for inspection.
• Windows x86 build is back!
• node_modules is now cached, saving 3min30s of build time and an additional 2min for Windows x86 builds.
• Cleaned up workflows.

[paulfitz]

Looking good overall, thanks @SleepyLeslie. Some questions.

[paulfitz]

Why is this check removed? It was checking something pretty subtle and weird but I see there are no comments as to what exactly :)

[paulfitz]

(Not against removing, just checking why)

[SleepyLeslie]

I looked at the code and couldn't understand why this check needs to be present. It was doing no useful work (basically if true; then echo ok; fi then repeat for another time), and there was no comment to explain its significance. I thought it was legacy leftovers not cleaned up. I double checked my CI and packaging runs and this step never failed (I would be surprised if it did). My guess is it might be checking whether the Windows association step worked (script is run with bash rather than powershell), and was likely added for debugging purposes, so I removed it. If anything breaks in the future I can always come back to it.
Please do share any extra context you have on what the problem was.

[paulfitz]

Yes, it was likely something like that, figuring out the association, and that a good version of bash was running. Doing bash-like stuff in windows was delicate, and I could easily imagine it breaking in future again. Still, since these checks have no comments, fine to remove them. Thanks for explaining.

[paulfitz]

Do you know what this was?

[SleepyLeslie]

According to this documentation, this option does nothing when oneClick is true (default, we didn't change it), so I removed it.

[SleepyLeslie]

FYI otherwise this option tells NSIS to skip the installation mode page and always installs Grist Desktop system-wide. This might not be a good idea in restricted environments where user does not have admin privileges, but we now have the portable build so this shouldn't matter much?

[paulfitz]

Pretty sure this was the hard-to-find solution to some gnarly problem, but we shouldn't keep it out of superstition. Someone will shout if it turns out to have some effect that was important, and then we can document the reason better next time :-). Cheers.

[paulfitz]

Thanks @SleepyLeslie ! Good to land.

[paulfitz]

Yes, it was likely something like that, figuring out the association, and that a good version of bash was running. Doing bash-like stuff in windows was delicate, and I could easily imagine it breaking in future again. Still, since these checks have no comments, fine to remove them. Thanks for explaining.

[paulfitz]

Pretty sure this was the hard-to-find solution to some gnarly problem, but we shouldn't keep it out of superstition. Someone will shout if it turns out to have some effect that was important, and then we can document the reason better next time :-). Cheers.