Corrupt snapshot prevents glusterd startup, no event API notification

[kohlschuetter]

An incomplete/corrupt snapshot entry under /var/lib/glusterd/snaps prevents glusterd from starting.
Instead, the snapshot should probably be marked "corrupt", deleted or moved elsewhere.

Steps to reproduce (tested with latest)

1. Create an invalid snapshot entry (an empty directory is sufficient)

mkdir /var/lib/glusterd/snaps/invalid
/etc/init.d/glusterd restart

Result:
glusterd start hangs

Output from /var/log/glusterfs/glusterd.log:

[2025-01-29 11:11:03.855392 +0000] E [MSGID: 106196] [glusterd-store.c:3969:glusterd_store_retrieve_snap] 0-management: Failed to update snapshot for invalid snap 
[2025-01-29 11:11:03.855399 +0000] E [MSGID: 106042] [glusterd-store.c:4134:glusterd_store_retrieve_snaps] 0-management: Unable to restore snapshot: invalid 
[2025-01-29 11:11:03.855416 +0000] E [MSGID: 101018] [xlator.c:643:xlator_init] 0-management: Initialization of volume failed. review your volfile again. [{name=management}] 
[2025-01-29 11:11:03.855423 +0000] E [MSGID: 101064] [graph.c:476:glusterfs_graph_init] 0-management: initializing translator failed 
[2025-01-29 11:11:03.855430 +0000] E [MSGID: 101174] [graph.c:825:glusterfs_graph_activate] 0-graph: init failed 

[kohlschuetter]

Notably, no event with that specific error ("Initialization of volume failed" / "Unable to restore snapshot") is sent via the Events API. I only see a PEER_CONNECT/PEER_DISCONNECT sequence.

We already have a few SNAPSHOT_*_FAILED events defined, but I guess something like SNAPSHOT_INIT_FAILED is what's missing here.