[Gap]: Document vMCP performance and sizing guidance

[yrobla]

What needs documentation?

Add documentation for vMCP performance characteristics and configuration options.

The following content should be added to the appropriate
documentation pages:

1. Sizing guidance (add to intro.mdx as deployment planning
   context)

2. Health check configuration (add to configuration.mdx or
   backend-discovery.mdx)

Context and references

No response

Use case

No response

Additional context

No response

[danbarr]

Preserving the original sizing/performance content from #499

Sizing guidance (assuming this is validated), add a version of this to the vMCP intro page:

## Performance considerations

### Recommended backend count

- **Small deployments**: 1-10 backends per VirtualMCPServer
- **Medium deployments**: 10-50 backends per VirtualMCPServer
- **Large deployments**: 50-100 backends per VirtualMCPServer

Beyond 100 backends, consider splitting into multiple VirtualMCPServer instances
with different groups for better performance and isolation.

### Resource usage

Approximate resource consumption per backend:

- **Memory**: ~5-10 MB per backend connection
- **CPU**: Minimal (~0.1m baseline + request handling)
- **Network**: Depends on tool call frequency

For 50 backends, expect ~250-500 MB memory baseline.

[danbarr]

Preserving the original health checks content from #499

Not directly related to backend discovery configuration, so I suggest adding to the main vMCP config guide.

### Health check configuration

vMCP performs health checks on backends using configurable intervals:

**Default settings:**

- **Check interval**: 30 seconds
- **Unhealthy threshold**: 3 consecutive failures before marking backend as
  unhealthy
- **Status reporting**: Status updates every 30 seconds

**Configure health checks** via `spec.config.operational.failureHandling`:

```yaml
spec:
  config:
    operational:
      failureHandling:
        healthCheckInterval: '10s' # Check every 10 seconds
        unhealthyThreshold: 5 # Require 5 failures before marking unhealthy
        statusReportingInterval: '15s' # Report status every 15 seconds
        partialFailureMode: 'best_effort' # Continue with healthy backends
        circuitBreaker:
          enabled: true
          failureThreshold: 10
          timeout: '120s'
```

**Configuration fields:**

| Field                             | Default  | Description                               |
| --------------------------------- | -------- | ----------------------------------------- |
| `healthCheckInterval`             | `"30s"`  | Time between health checks                |
| `unhealthyThreshold`              | `3`      | Failures before marking unhealthy         |
| `statusReportingInterval`         | `"30s"`  | Status update frequency                   |
| `partialFailureMode`              | `"fail"` | Behavior when some backends fail          |
| `circuitBreaker.enabled`          | `false`  | Enable circuit breaker pattern            |
| `circuitBreaker.failureThreshold` | `5`      | Failures before opening circuit           |
| `circuitBreaker.timeout`          | `"60s"`  | Wait time before retrying failed backends |

**Partial failure modes:**

- `fail`: vMCP returns error if any backend is unhealthy (strict)
- `best_effort`: vMCP continues with healthy backends only (tolerant)

Health checks use the backend's MCP `initialize` method to verify connectivity.