Enhance exponential backoff retries for HTTP requests to work for connectivity issues

[theJC]

Description

In #693 HTTP retries with backoff were added, but appear to only be for when a response is received and its an internal error (based on the release notes, haven't found any other documentation yet)

Please expand this capability to retry on connection failures or timeouts.

Bonus points/stretch goal would be to add command line arguments to control the number of requests and timing of the backoff strategy used.

This will make everyone's use of rover in CICD pipelines much more hardened and less likely to fail. In our organization we typically get several messages a week in our support channels where teams have had their CICD pipeline fail on them because of rover having connect connection or timeout issues when checking our publishing their schema.

This past week, the incident: https://status.apollographql.com/incidents/4xcz62jzl7s9 was responsible for a spate of customers pipelines failing.

[ptondereau]

@abernix I would like to pick the issue, but I'm seeing that you're assigned to it.
May I propose something?

[EverlastingBugstopper]

Hey @ptondereau - we'd love your proposal! Sorry your comment was lost in the fray.

[ptondereau]

Hey @ptondereau - we'd love your proposal! Sorry your comment was lost in the fray.

I'm very busy since I've found a new job, but I'll try to come with a PR quickly!

[EverlastingBugstopper]

OK! If it turns out you don't feel like working on this then don't feel any obligation to, just wanted to let you know that we'd accept a PR if you decided it was worth your time 😄