[Extension] Refresh hosts/sources on Mount failures instead of using cached creds

[prafgup]

Issue #, if available:

Exploration based on this - awslabs#1927 (comment)

Mount is called in two scenarios:

  1. Prepare (line 399) — Called every time containerd pulls/creates a container snapshot. This happens on every ctr run / pod creation, not just the first pull. If the image is
already in containerd's content store but the SOCI lazy-loading snapshots need to be set up (e.g., container restart, new pod from same image), Prepare → prepareRemoteSnapshot → Mount runs again.
  2. Restart recovery (line 1146) — When the snapshotter restarts, it re-prepares all existing remote snapshots via prepareRemoteSnapshot.

Description of changes:

Testing performed:

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.