Ignore unexpected metric messages

[bmario]

In conjunction with metricq/metricq-manager#39, we could (and should) ignore unexpected messages, as the manager and the DB are - in theory - on the same page. Thus the configuration should always be okay. And under normal operation, It works fine.

However, there is this check.

Image the following situation:

• You remove metric M from the DB config
• You restart the DB to apply the change, as you need to
• During the restart, the manager reconfigures the queue, but there is at least one message for metric M in the queue
• During the DB restart, the DB receives this one message for M
• M is no longer in the config for the HTA::Directory, so that will raise an exception, which gets caught and ignored. Fine, but this message also creates an entry in the mapped_metrics_ entry because, for the wildcard DB config, we need to add mappings on the fly.
• Now, you realize, you didn't mean to remove the metric M
• So you add back the config entry for M
• You send a reconfigure RPC to the DB
• The DB performs the reconfiguration, but look again at the linked check from above. The metric M already has an entry in the mapped_metrics_ member. So its configuration never gets added to the HTA::Directory.
• Now, the messages for metric M will get ignored, as there is no configuration for it in the HTA::Directory

TL;DR Meme: