Serialize MDC under top level

[felixbarny]

Makes it easier to set ECS fields via MDC.
Previously, you'd have to configure topLevelLabels for each ECS field added via MDC.
Now, if you want to have a field nested under labels, you have to explicitly prefix it with labels..

This increases the chance of conflicts, for example when the user adds an @timestamp MDC value. But it also gets out of the way for legitimate use cases.

Having to whitelist all MDC values that should be serialized at the top level via topLevelLabels is cumbersome and feels inconsistent. I think it's overly protective.

[felixbarny]

@pgomulka would this break anything in Elasticsearch?

[xeraa]

Took a quick look and I think this makes a lot of sense (also in documenting how to use this)

[pgomulka]

@pgomulka would this break anything in Elasticsearch?

@felixbarny ES is not using MDC, so should be safe.

[lanore78]

With this change, I'm having a problem with collecting log with Filebeat.

It's because some library produce 'url' MDC which is one of ECS key, and the formate is not compatible with ECS(or Elasticsearch)

Here is example value of 'url' MDC, and I got error when I collect this log with Filebeat.
Is there any good idea to solve this problem?

{ ...
"url" : "jdbc:sqlserver://xxx.xxx.xxx.xxx:61051;authenticationScheme=nativeAuthentication;xopenStates=false;sendTimeAsDatetime=true;trustServerCertificate=false;sendStringParametersAsUnicode=false;selectMethod=direct;responseBuffering=adaptive;packetSize=8000;multiSubnetFailover=false;loginTimeout=20;lockTimeout=-1;lastUpdateCount=true;encrypt=false;disableStatementPooling=true;databaseName=USER;applicationName=Microsoft JDBC Driver for SQL Server;applicationIntent=readwrite",
...
}