feat(feed): use exponential backoff for broken feeds

[jvoisin]

Instead of having a hard limit of 3 (regular) tries for feeds with errors, use an exponential backoff by default, with an upper-limit of one week.

See #3334 for context/discussions.

[fguillot]

Where are the unit tests?

The config options SCHEDULER_*_MIN/MAX_INTERVAL probably needs to be taken into consideration, and/or new config options needs to be added for the backoff delay and upper limit.

[jvoisin]

Where are the unit tests?

They were obviously replaced with unwarranted cockiness, thinking that such trivial feature wouldn't warrant them, which is, in retrospect, completely wrong.

I'm switching this PR to draft, and will mark it as ready once comprehensive testing is added.

[jvoisin]

new config options needs to be added for the backoff delay and upper limit.

The subtitle on https://miniflux.app states that "Miniflux is a minimalist and opinionated feed reader." Adding more options to the ~150 possible ones feels anything be minimalist nor opinionated. And even if the slogan were "Maximalist and ultra-personalisable", I can't think of valid reasons to make the low and upper bound of the backoff configurable.

For example, there is no point trying to refresh a feed forever if the website no longer exists.

I disagree: sometimes websites go down (eg. Low→Tech when there is no Sun, or the myriad of self-hosted blogs that are going down when their owner/hoster is moving to a new place, or when my fiber at home gets cut and miniflux fails to refresh feeds for a week because there is no internet, …)

[jvoisin]

The linter being unhappy is unrelated, and should be fixed by #3870