Log durations for failing GraphQL-queries
What does this MR do and why?
If the query were to break by a Runtime error, even if it's handled
later on in the case of an GraphQL::Error
subclass.
This also ensures that we don't record an apdex for those failing queries, those will have to be covered by other metrics. This was already the case, but now there is a spec proving that.
For #329288 (closed)
Edited by Bob Van Landuyt