feat: add max_retry_backoff option to job runner

[sirewix]

A solution to #19 based on StructionSite@0489b51

[GoldenChrysus]

This looks very useful. I have a sentinel job that performs some cleanup once a channel is empty. It does that with checkpoints but a long-running sentinel will end up with an absurdly high backoff as there is no option for linear retries. Looks like this PR would allow me effectively avoid using exponential backoff in the first place.

I wonder if @Diggsey has abandoned this project? PRs have been sitting for a long time despite very active GH activity...