tsdump performs strict checks that verify that it has data for exactly the stores it expects to see according to the node->store mapping. It can happen that there legitimately isn't data there, for example when nodes are down during the tsdump window. We still want the strict checks, but now tsdump will should defer returning an error and allow restarting with a - file to display the data anyway.

Cluster settings are should be set via SET CLUSTER SETTING as opposed to overriding settings directly.

Jira issue: CRDB-12908