fix(ci): add http retries and bigger timeouts

[lidel]

Aims to close #532 by bumping timeout from 1m:

TODO

• add retry to curl that downloads go-ipfs binary
• update to latest http client, which no longer have client-side timeouts nvm, 🕳️ 🐇, not touching this before xmas
• instead, set global timeout of 45m for all ipfs-http-client operations during publish step

[github-actions]

This change produced no new differences in built artifacts.

[aschmahmann]

update to latest http client, which no longer have client-side timeouts

What does this mean it'll just hang forever unless you tell it explicitly how long to timeout after? It's fine to increase the timeout here, but it'd be nice if we could error here and kill the build instead of rebuilding any binaries where we failed to do the ipfs cat.

This behavior of rebuilding on failure is also a little bit dangerous in the sense that it seems like we could end up pushing binary changes to dist.ipfs.io for old binaries accidentally.

[lidel]

I meant implicit timeout in [email protected] (code here).
iiuc it will timeout after 20 minutes even if go-ipfs default is to wait for data forever.

That being said, it does not seem to be low enough (20m) to produce issues we experience (timeout <1m).
It starts to look like an issue with github infra itself: docker just failed to start witht he same context deadline exceeded error 🙃

Pushed some explicit retries and timeouts and restarted the job. If it fails again, let's wait for github to calm down until tomorrow.

[github-actions]

This change produced no new differences in built artifacts.

[github-actions]

This change produced no new differences in built artifacts.

[github-actions]

This change produced no new differences in built artifacts.

[github-actions]

This change produced no new differences in built artifacts.

[github-actions]

This change produced no new differences in built artifacts.