Reset retry backoff after max retry interval

[keelerm84]

According to our streaming spec, default retry behavior should continue with a backoff calculation until 60 seconds of continuous failures have occurred, after which we can reset the retry interval.

While working on this task, I extracted the retry strategy into a separate struct. This keeps the client a little cleaner, allows us to actually unit test the retry policy, and will hopefully more easily allow us to control the retry behavior for future unit tests.

[shortcut-integration[bot]]

This pull request has been linked to Shortcut Story #174725: Only reset stream delay after 60 seconds.

[keelerm84]

I'm pretty out-of-context - am I reading it right that there's no jitter (in the previous implementation or this one)?

Correct. I am doing that in the next ticket.