sambacc: avoid logging an error if cluster is being torn down

Saw this in a ceph teuthology run:
```
  2024-08-20 20:39:57,289: DEBUG: Creating RADOS connection
 2024-08-20 20:39:57,333: INFO: cluster meta content changed
 2024-08-20 20:39:57,333: DEBUG: cluster meta: previous={'nodes':
[{'pnn': 0, 'identity': 'smb.adctdb1.0.0.ceph0.kdlxgn', 'node':
'192.168.76.200', 'state': 'ready'}, {'pnn': 1, 'identity':
'smb.adctdb1.1.0.ceph1.ngbqkk', 'node': '192.168.76.201', 'state':
'ready'}, {'pnn': 2, 'identity': 'smb.adctdb1.2.0.ceph2.rhmqnu', 'node':
'192.168.76.202', 'state': 'ready'}], '_source': 'cephadm'} current={}
 2024-08-20 20:39:57,333: ERROR: error during ctdb_monitor_nodes: max()
arg is an empty sequence, count=0
 Traceback (most recent call last):
  File "/usr/lib/python3.9/site-packages/sambacc/commands/ctdb.py", line
479, in catch
    yield
  File "/usr/lib/python3.9/site-packages/sambacc/commands/ctdb.py", line
360, in ctdb_monitor_nodes
    ctdb.monitor_cluster_meta_changes(
  File "/usr/lib/python3.9/site-packages/sambacc/ctdb.py", line 561, in
monitor_cluster_meta_changes
    expected_nodes = _cluster_meta_to_ctdb_nodes(
  File "/usr/lib/python3.9/site-packages/sambacc/ctdb.py", line 506, in
_cluster_meta_to_ctdb_nodes
    pnn_max = max(n["pnn"] for n in nodes) + 1  # pnn is zero indexed
 ValueError: max() arg is an empty sequence
```

I could see from the ceph logs the smb cluster was being removed right
around this time. If we had nodes and they suddenly vanish we're likely
in the process of getting removed and we raced a tad with cephadm
removing services while the smb mgr module was removing the contents
of the .smb pool.

Signed-off-by: John Mulligan <[email protected]>

Author: phlogistonjohn

sambacc/ctdb.py

@@ -551,6 +551,13 @@ def monitor_cluster_meta_changes(
         if curr_meta == prev_meta:
             _logger.debug("cluster meta content unchanged: %r", curr_meta)
             continue
+        if len(prev_meta) > 0 and len(curr_meta) == 0:
+            # cluster is possibly (probably?) being destroyed.
+            # Return from this loop and let the command-level loop decide if
+            # this function needs to be restarted or not. There's a chance this
+            # process will be terminated very soon anyway.
+            _logger.warning("no current nodes available")
+            return
         _logger.info("cluster meta content changed")
         _logger.debug(
             "cluster meta: previous=%r current=%r", prev_meta, curr_meta