Add `extropy config test` command to verify model connectivity

[DeveshParagiri]

Problem

There's no way to verify that configured models are working before running a pipeline. You only find out when a command fails mid-run, potentially after spending time/money on partial execution.

Proposed Solution

Add extropy config test command that:

1. Pings each configured model (pipeline fast, pipeline strong, simulation fast, simulation strong) with a minimal request
2. Reports success/failure for each
3. Shows latency and any error messages

Example Output

$ extropy config test

Testing model connectivity...

Pipeline models:
  ✓ anthropic/claude-sonnet-4-6 (fast) — 342ms
  ✓ anthropic/claude-sonnet-4-6 (strong) — 358ms

Simulation models:
  ✓ azure/gpt-5-mini (fast) — 128ms
  ✓ azure/gpt-5-mini (strong) — 131ms

All models responding.

$ extropy config test

Testing model connectivity...

Pipeline models:
  ✓ anthropic/claude-sonnet-4-6 (fast) — 342ms
  ✗ anthropic/claude-sonnet-4-6 (strong) — Error: Invalid API key

Simulation models:
  ✓ azure/gpt-5-mini (fast) — 128ms
  ✓ azure/gpt-5-mini (strong) — 131ms

1 model failed. Check API keys and model availability.

Implementation Notes

• Use the existing LLM client infrastructure
• Send a trivial completion request (e.g., "Say 'ok'" with max_tokens=5)
• Consider adding --model <model> flag to test a specific model
• Exit code 0 if all pass, non-zero if any fail