meta tags: avoid frequent write locks

[shanson7]

This is a branch for a concept change. Right now we need to frequently relinquish the write lock to avoid deadlock scenarios when adding/deleting series within the meta tags component.

As discussed before acquiring the write lock can be very expensive. This change is designed to reduce the number of times we need to acquire the write lock within batching operations.

Open question: is calling getOrgMetaTagIndex under lock a potential deadlock?

[replay]

Open question: is calling getOrgMetaTagIndex under lock a potential deadlock?

I don't think so. .getorgMetaTagIndex() only requires the lock on metaTagIdx. This lock only gets acquired by .getOrgMetaTagIndex() and .stop().

getOrgMetaTagIndex() is very straight-forward, it never waits for something else, it will always releases the lock very quickly again.

.stop() is more complex because it waits for the enrichers to shutdown, this shutdown happens by sending a shutdown command through the enricher queue, so if that queue were blocked by the lock on the main index then this would result in a dead lock. but since #1834 it shouldn't be possible anymore for the enricher queue to get blocked on waiting for the main index lock.

[replay]

LGTM

[shanson7]

Could someone merge or take a look?