[Feature] Dynamically load IaaS credentials during runtime

[timuthy]

Feature (What you would like to be added):
The etcd-backup-restore process reads IaaS credentials for access cloud provider storage when it creates the Snapstore. Most of the times the credentials change, e.g. due to secret rotation, the process needs to be restarted (depending on Snapstore provider).

To circumvent a required restart, etcd-backup-restore should re-read the credentials every time it performs any action on the bucket.

Motivation (Why is this needed?):
Dynamic credential retrieval prevents avoidable pod/container restarts which always involves a downtime of the etcd cluster if only one replica is involved.

Approach/Hint to the implement solution (optional):
It's important to use a file-based approach for every supported provider (already true for GCP today) rather than storing the access information in env vars. etcd-backup-restore is mostly deployed as a sidecar and content refreshing is only supported if Kubernetes secrets are mounted into the container (see https://kubernetes.io/docs/concepts/configuration/secret/#mounted-secrets-are-updated-automatically).

[timuthy]

cc @ishan16696 @abdasgupta @mliepold

[ishan16696]

etcd-backup-restore should re-read the credentials every time it performs any action on the bucket.

Currently we just create snapstore object once and used the same obj throughout to access object bucket.
To do this there is no other option but to create/update the snapstore object every time backup sidecar tries to take the snapshot(full or delta).

[ishan16696]

To Dynamically load IaaS credentials during runtime:

• To circumvent a required restart, etcd-backup-restore should recreate/update the snapstore object every time it performs any action on the bucket.
• Mount Kubernetes secrets into the container:
  • Google cloud storage(GCS)
  • AWS s3
  • Azure Blob Storage
  • Alicloud Object storage
  • Swift (Openstack)
• Open another similar PR for master branch