core/zero: fix ticker usage

[calebdoxsey]

Summary

In several places we use a time.Ticker based on a backoff.Backoff. We create a new time.Ticker that fires immediately by using a short duration. Unfortunately this causes excessive CPU usage. There are several ways to fix this:

1. We can use the backoff.NewTicker function, which returns a "Ticker" that behaves as expected by using a time.Timer instead. This approach does not work if we need to call Reset on the Backoff.
2. We can use the backoff.Retry function with a backoff.WithContext. We then don't need a timer or ticker.
3. Instead of a time.Ticker we can use a time.Timer set to the backoff interval. Since a time.Timer is allowed to be set to 0 with no downside, this can handle the case when we want to fire immediately. The benefit over a time.After is we can call .Stop to collect resources if another case in the select fires.

Option (3) was used for the retry.Retry function because the logic around resetting is incompatible with (1) and (2).

Related issues

Fixes https://github.com/pomerium/internal/issues/1727

Checklist

• reference any related issues
• updated docs
• updated unit tests
• updated UPGRADING.md
• add appropriate tag (improvement / bug / etc)
• ready for review

[calebdoxsey]

I should also mention that a call to Reset on Timer can only be done if the Timer has already fired, which prevents some ways of solving this problem.

[coveralls]

coverage: 56.456% (+0.03%) from 56.426%
when pulling bec9b5f on cdoxsey/fix-ticker
into 0d2705b on main.

[calebdoxsey]

Sorry forgot I wasn't supposed to close PRs. Re-opening.