Why do we separate the UserService and UserServiceImpl?

I think it's to go along best practices, to code against an interface. Imagine the following two scenarios:

1. I put a pencil on a table and tell you to write something down.
2. I put a pen on a table and tell you to write something down.

You would be able to write something down because both pen and pencil both "implement" the action "write" even though in different ways. You wrote with whatever was on the table, it didn't really matter what, because you know both pen/pencil can do it.

Now imagine this scenario:

1. I put a pencil on a table and tell you to write something down a pen. You won't be able to do it because there is no pen. Because my instructions were tied to the concrete implementation of "write".

The same goes with the Service (write something down). The actions are geared towards it, so it doesn't matter which ServiceImpl (pen/pencil) it's being used. It allows for it to be changed in the future without breaking the code.

The benefit is that it allows to change concrete implementations without breaking existing code, it becomes flexible and extensible.

Hope it made more or less sense.