doc: distributed computing with ray

Use cases, pain points, and background
The Ray documentation in reference/faq.md has several issues:

1. Wrong location — Buried in FAQ, hard to discover for users hitting scale issues
2. Unclear audience — Mixes what's automatic (Ray init) with what's optional (using @ray.remote in your code)
3. Missing decision criteria — Doesn't explain when you need Ray
4. Generic example — Shows standalone function, not integrated with resource server patterns

Description:
Relocate documentation from FAQ into standalone docs page and improve the content for clarity

Design:

# Distributed Computing with Ray

## Overview
<!-- What Ray does in NeMo Gym, that it's auto-initialized -->

## When You Need Ray
<!-- Decision criteria: CPU-bound verification, 1000s of concurrent requests -->

## Using Ray in Your Code
<!-- How to add @ray.remote to your functions -->

## Configuration  
<!-- ray_head_node_address for training framework integration -->

Out of scope:

• General Ray tutorials or Ray cluster administration (link to Ray docs instead)
• Changes to Ray initialization logic in NeMo Gym code
• Performance benchmarking or profiling guidance (covered in separate profiling docs)
• Troubleshooting Ray-specific errors (e.g., serialization issues, memory limits)
• Multi-node cluster setup instructions (training framework responsibility)

Acceptance Criteria:

• New docs page created at docs/infrastructure/distributed-computing.md
• Page covers all 4 sections: Overview, When You Need Ray, Using Ray, Configuration
• Example code shows integration with a resource server verify() method (not standalone function)
• Decision criteria section includes concrete thresholds (e.g., "CPU-bound tasks taking >Xms at >Y concurrent requests")
• FAQ entry removed entirely
• Toctree updated in docs/index.md with new Infrastructure section
• Docs build successfully (make html in docs/)

Start by reading the Ray-related content in reference/faq.md, then inspect docs/index.md and nearby pages under docs/infrastructure/. Create docs/infrastructure/distributed-computing.md with the four specified sections and resource-server verify() example, remove the FAQ entry, update the toctree, and run make html in docs/ to confirm the build succeeds.