[RFE] Improve the node_lock

[phvalguima]

Currently, we have a node lock that has proven to work really well for 3+ node deployments.

However, as we scale below 3, the lock will eventually switch its backend from the opensearch index .charm-node-lock to the peer relation databag.

This is really confusing, as the lock code itself gets filled with if/elif/else options to handle these cases.

My proposal is to simplify this lock:

1. Remove the peer relation lock: the opensearch index is enough when you have >=1 online nodes; and handling the case of 1 single node is trivial
2. Make the decision factors of how the lock will be used (e.g. should we use node_lock, peer_lock, etc) really visible in the acquired method.

[github-actions]

https://warthogs.atlassian.net/browse/DPE-5125