It’s historical more than anything else. The org.junit
package was first and had a mixture of the API and implementation classes for running the tests and test cases.
For the JUnit 5 api, the classes were split across different packages, so that when you have the annotations they can be included separately for your runtime and the test execution engine is in a different package.
To make things easy, the classes kept their same names so you just needed to change the package included on the class path.
So you can use either, but try and use one package consistently.
The other observation is that the reason for keeping the packages separate is so hat you could have a project which consisted of some old JUnit libraries, and some new libraries. So you wouldn't need to refactor everything in one go to migrate.
https://junit.org/junit5/docs/current/user-guide/#migrating-from-junit4
PS if you want to ignore the older JUnit packages you can add them to the step filters
in Eclipse's preferences, and then Eclipse won’t suggest that package.