[MLA-1880] Raycast sensor interface improvements

[chriselion]

Proposed change(s)

Public-facing change:

• Adds RayPerceptionSensor.RayPerceptionOutput, which contains the most recent ray hit information. This can be used to for additional game logic as needed.
• Adds new fields to RayOutput: StartPositionWorld, EndPositionWorld, ScaledRayLength, and ScaledCastRadius.

Internal changes:

• Get rid of DebugDisplayInfo - all of its info is now on RayOutput, except for the frame that the results were queried, which is now stored as m_DebugLastFrameCount
• Raycasts now happen in the Update() call, which always happens before the Write().

Useful links (Github issues, JIRA tickets, ML-Agents forum threads etc.)

Based on feedback from several forum posts:

Forum threads:

• "i want to do an if statement inside my agent script to say if the ray cast hits an object with the tag enemy then do something"
• "does the ray perception 3d component get the distance to the object it impacted? I need my agents to get the distance to that object"
• "we have to reset agent if it follows wrong raycast and crash, or we want to add reward if it follows correct raycast."
• "I am trying to teach an agent to find a target but once it has "seen" the target, I want to give it a reward."
• "I was wondering, if there is an easy way to draw the Debug gizmos of the Ray Perception Sensor ingame for demonstration purposes?"
• "I'd like to get a value of distance between the agent and obstacles with using Ray Perception Sensor."
• "I use three Ray Perception Sensor Components for measuring distances from the agent to some objects, and I'd like to get all values (HasHit, HitTaggedObject, HitTagIndex and HitFraction) from all components."

Types of change(s)

• New feature
• Code refactor

Checklist

• Added tests that prove my fix is effective or that my feature works
• Updated the changelog (if applicable)

Other comments

[dongruoping]

Looks good to me. Verified that the gizmos work fine in editor.