configurable network retries

[cgwalters]

We should support configuration for network errors.

Related:

• upgrade/switch: Support trickle-fetch #1375
• Support retries on commands containers/skopeo#918
• https://issues.redhat.com/browse/OCPBUGS-43406

Especially in some scenarios, higher level software may want to retry a lot (e.g. potentially minutes) for network errors.

Bikeshed: Unlike podman/skopeo I'd say --network-retry instead of just --retry to be crystal clear what it does/doesn't do.

I think we could add a generic NetworkRetries struct that gets added to upgrade/switch/install.

NOTE this depends on #1424 because it effectively pulls in the retryable error code from https://github.com/containers/skopeo/blob/main/docs-experimental/skopeo-experimental-image-proxy.1.md