docs: add crisis learning examples to ai-provider-agnosticism principle

## Summary
Enhance the ai-provider-agnosticism principle with real crisis learning examples from this week.

## Concrete Examples to Add

### Claude 500 Errors → Amazon Q Resilience
```bash
# Monday: Claude down with 500 errors
$ claude
Error: Service unavailable

# Solution: Switch to Amazon Q instantly
$ q
# Same MCP configs, same knowledge/, zero downtime
```

### The Math of Crisis
- **Without agnosticism**: 4 hours downtime = 0 productivity
- **With agnosticism**: 5 second switch = continuous flow
- **ROI**: 2880x productivity preservation (4 hours vs 5 seconds)

### Implementation Speed
Document how we went from crisis to solution in hours:
1. Claude fails → identify pattern
2. Create `q` alias with MCP import
3. Test identical workflows
4. Document and ship

## Why This Matters
- Shows principle validated under fire
- Provides template for future crises
- Demonstrates "never let a crisis go to waste"
- Real numbers make the case for provider diversity

Principle: ai-provider-agnosticism
Principle: systems-stewardship

Locate the documentation for the ai-provider-agnosticism principle and review how it relates to systems-stewardship. Add the Claude 500-error and Amazon Q example, crisis timing and ROI figures, and the four-step implementation sequence; done means these real-world examples are documented clearly under the relevant principles.

Written by the indexing model from the issue text.