[object_store] Retry requests on connection error

[kindly]

Describe the bug

When doing mulitipart uploads to s3, for larger files, I was getting lots of reqwest/hyper errors, e.g broken pipe, tls handshake errors and channel closed, meaning some large files become almost impossible to upload. This could be due to flaky internet or slow processing of data being uploaded.

To Reproduce

Hard to reproduce as they were intermittent failures and required large files or delays between requests.

Expected behaviour

Ability for the put requests/multiplart uploads to be retried for more failure cases.

Additional context

Pull request #4120 does retries for more cases, which fixes these issues for me.

[tustvold]

label_issue.py automatically added labels {'object-store'} from #4120