Add `--no-meshes` and `--no-textures` arguments to reduce RAM usage

[lukicdarkoo]

In Reinforcement Learning applications, devices derived from AbstractCamera are usually not used. Therefore, we can disable meshes and textures to reduce the RAM footprint.

I have just done some tests with UR5e in an empty world:

• 850MB with initial setup,
• 400MB with all graphics enchantments turned off,
• 177MB with meshes removed, and
• 126MB with textures disabled.

We can add --no-meshes and --no-textures CLI arguments to disable all meshes and textures in the simulation.

---

If you need to do something similar now then you can add a boolean PROTO field and using Lua add conditional includes for meshes and textures.

[Simon-Steinmann]

Scenario: No camera based device in simulation suggestion: Turning off rendering should turn the graphics enchantments off, freeing up RAM