Adds Anymal-C navigation rsl_rl example with obstacle and camera in the scene

[KingsleyLiu-NV]

Description

Please include a summary of the change and which issue is fixed. Please also include relevant motivation and context.
List any dependencies that are required for this change.

This PR adds the anymal-c navigation rsl_rl example with obstacle and camera in the scene:

• Extends the rsl_rl cfg and exporter to support convolution networks
• Extends the anymal-c navigation environment to have rigid object obstacle and tiled camera sensor
  • Adds the event reset function to sample the position in a sector of ring area
  • Adds the observation function for the flattened normalized rgb data
  • Adds the pose3 command resample function to determine the target based on the obstacle position
• Tune the RL reward terms to make the training converge

Type of change

• New feature (non-breaking change which adds functionality)
• This change requires a documentation update

It can be run with the following command:

./isaaclab.sh -p scripts/reinforcement_learning/rsl_rl/train.py --task Isaac-Navigation-Flat-Obstacle-Anymal-C-v0 --num_envs 1024 --max_iterations 1501 --enable_cameras --headless

./isaaclab.sh -p scripts/reinforcement_learning/rsl_rl/play.py --task Isaac-Navigation-Flat-Obstacle-Anymal-C-Play-v0 --enable_cameras --num_envs 1 --video --video_length 120 --headless

Screenshots

The training results are as below:

Play with model_1500.pt:

play-120step.mp4