I don't think the github project approach is going to work. What I'm suggesting is more than just provide some kind of commonly used library that does annotation scanning. I've already got a commonly used library that does annotation scanning - it's Eclipse Jetty. I could spin out our code for annotation scanning to github, but it will still just be one of many many implementations with nothing to recommend it over any of the many other implementations.
This effort needs not to be code driven, but requirement/specification driven. As the links given indicate, it should have been functionality that was included in Java 9, but it didn't happen. Unless it does happen, then all the various projects are going to separately implement, with many differences and incorrect implementations and we are all going to suffer through lots of instability as java 10 and 11 are released on the proposed 6 monthly schedule. So there needs to be a collaborative effort that has a bit more gravitas than some random github project.
Note that Eclipse Jetty as a servlet container is not limited to enterprise, yet we follow a specification that is driven by the EE effort, originally at the JCP and now here. So I don't really see why EE4J cannot consider coordinating this?