withElasticSearchLogger continues past `(Logger -> IO r)` action when EL is unreachable

[jonathanjouty]

As per title, if your Elasticsearch cluster is unreachable or cannot accept messages for whatever reason, withElasticSearchLogger will keep going even though the (Logger -> IO r) action completed.

This is not completely counter to the documentation, which states that:

Create an elasticSearchLogger for the duration of the given action, and shut it down afterwards, making sure that all buffered messages are actually written to the Elasticsearch store.

The question is more about how this should work—I would prefer if it actually gave up after some time... since Elasticsearch can easily be unavailable.

Maybe an extra timeout parameter?

cc/ @arybczak @23Skidoo @phadej

[jonathanjouty]

Ping @arybczak @phadej @23Skidoo
Any preferred solution before we start working on this?