fix jmx service checks

[arbll]

What does this PR do

JMXfetch used the name of the configuration instead of the check name in the configuration to send service checks.
The recent refactor on datadog-agent that allowed having multiple configs for jmx by giving name different name exposed the issue and broke service checks for jmx. This fixes it.

[olivielpeau]

LGTM, a couple of questions though:

• I understand this code path is only active when JMXFetch is used with the Agent 6 (and not with the Agent 5)?
• Is the check name here only used for the service check name, or should we look out for other uses of it in tags/JMXFetch's logic?

[arbll]

• Yes, since they both have their own way of fetching configs this isn't too surprising. Agent5 allows for multiple config with the same name since it uses an Iterator. Agent6 needs to to have different names for configuration since we map the JSON to a dictionary.
• I'll double check everything but this PR changes the check name at the instantiation of the check. It should replicate over everything using the check name.

[truthbk]

This looks good to me, we're now using the right check name as submitted by the agent. This doesn't just affect service checks, but all metrics too.

The instance: tag had been broken in the previous agent release as we use the check name to build the instance tag. Unfortunately, this fix will cause churn to customers on 6.3.0. The fix will naturally also affect the service check as intended.