chore: timeouts on version discovery

[dpc]

When working locally I've noticed test failing due to too short api discovery timeout.

Parametrize timeout of api version discovery using env variable. Set it to low value only when testing offline nodes.

[maan2003]

fedimint-client should not use environment variables.
libraries using env variables is super weird.
we should move env variables checking to fedimint-cli

[elsirion]

The problem with that is getting the setting deep into the client. Eventually we'll need a general solution for things like this to also override the default esplora server etc., but for now this seems like the only reasonable-effort way to adjust the timeout in tests without breaking clients on slow connections.

[dpc]

fedimint-client should not use environment variables.
libraries using env variables is super weird.
we should move env variables checking to fedimint-cli

I agree, but it isn't a big crime and that can be a follow-up. We'll have some builder pattern on Client where we can set this value, and read it from variable in fedimint-cli. But this PR is better than hardcoding a super-short timeout, which will break production cases, and all in all it isn't too bad and seems perfectly landable. @maan2003

[bradleystachurski]

@dpc Thanks for tackling this. I think we can close #4398, but feel free to keep open if you want to refactor the misdemeanor in a followup.

@maan2003 I share your discomfort with using env vars to trigger different logic in prod vs tests. I'm approving since I prefer the tradeoff to fix the broken behavior in prod first, which can be refactored in a fast follow.

[dpc]

copying path '/nix/store/mx1mwi6bfzjsblhw1m95qs1zwb0l382c-ndk-bundle-23.0.7344513-rc4' from 'https://fedimint.cachix.org/'...
Error: The operation was canceled.

@bradleystachurski Seems like the nix stale timeout you landed recently also didn't help? :(

https://github.com/fedimint/fedimint/actions/runs/8181905886/job/22372424769

It's always the ndk-bundle (which is probably relatively large), so maybe it's not completely stuck but just sometimes get very slow?

[bradleystachurski]

@dpc Should we try temporarily bumping the cross compile timeout?

[dpc]

@dpc Should we try temporarily bumping the cross compile timeout?

Worth giving a shot, I guess. Can you make a PR?

[bradleystachurski]

#4487