[Ray Data] Improve the design and documentation of _MapWorker

[marwan116]

Description

Background: _MapWorker are actors created by ActorPoolMapOperator, which is used in the case where users call map-like APIs with a callable class UDF.

Use case

I have a ray cluster that is stuck attempting to schedule a _MapWorker. This is executing a complex ray data pipeline which contains more than one UDF, so a few problems arise:

1. There is no way to tell which of the map UDF operations is stuck given it is only a generic _MapWorker and the user has to resolve the "optimal ray data DAG execution" in their head to guess what might be the issue.
2. Currently one has to be knowledgeable of the ray data implementation specifics to understand what _MapWorker does given the name on its own is not self-explicable

See the attached screenshot for more details

Improvement suggestions:

• Update the ray data docs to explain what the _MapWorker construct or whatever we end up naming it does, and what does its methods like __init__ andget_location do...
• Rename _MapWorker to something more public facing like LaunchWorker
• Always show the LaunchWorker under a subgroup of the MapWorker it is trying to create