docs: add model registry integration guide for operators (#1042)

## Summary

- Adds `docs/operator-guides/integrations/model-registry.md`
- Explains the `ModelVersion` CRD — what fields it carries, how artifact
paths are structured in S3, and how to query registry state via
`kubectl` and the `ma` CLI
- Covers downstream integrations: Michelangelo's own InferenceServer,
external serving infrastructure reading directly from S3, and syncing to
an external model registry via K8s watch
- CI/CD gating pattern: `kubectl wait` for `ModelVersion` Ready
condition before promoting to production serving, with a GitHub Actions
example
- RBAC section for granting read access to model registry resources
- Retention / cleanup guidance for S3 artifact lifecycle management

## Why

Operators need to understand how registered models flow out of
Michelangelo — into serving, into external governance registries, and
into promotion pipelines. This is the second integration guide in the
new `integrations/` directory, complementing the experiment tracking
guide (PR #1041).

## Test plan

- [ ] Verify all internal links resolve
- [ ] Confirm YAML snippets (ModelVersion, InferenceServer, RBAC) are
syntactically valid
- [ ] Confirm `jq` and `kubectl` commands in bash blocks are correct
- [ ] Verify no specific external registry vendor is named or implied

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Austin Greco <austingreco@gmail.com>
Co-authored-by: Austin Greco <agreco@uber.com>

Author: 3 people

docs/operator-guides/index.md

@@ -49,6 +49,7 @@ Michelangelo is designed to be adopted alongside existing ML infrastructure. The
 
 | Guide | Description |
 |-------|-------------|
+| [Model Registry](integrations/model-registry.md) | Operate Michelangelo's built-in model registry, configure storage and RBAC, and integrate with serving and CI/CD |
 | [Experiment Tracking](integrations/experiment-tracking.md) | Connect an external experiment tracking server to Michelangelo task pods |
 | [Custom Serving Backend](serving/integrate-custom-backend.md) | Add support for any inference framework — Triton, vLLM, TensorRT-LLM, or your own |
 | [Custom Job Scheduler](jobs/extend-michelangelo-batch-job-scheduler-system.md) | Replace or extend the job scheduler — integrate Kueue, Volcano, or a custom assignment strategy |