For XML metadata extraction you will still use the SDK (for more information, see the Alfresco In-Process SDK documentation) and a JAR project applied to the Repository (i.e. alfresco.war).
To change the configuration for the majority of the metadata extractors you will have to generate a new Transform Core AIO Docker image with the new configuration. Another option would be to create a new separate T-Engine that has a higher priority (lower number) for this metadata extraction. That way you can still use the standard T-Engine and the new one from for this one special case.