feat(guardrail): support streaming output rewrite

[coderabbitai]

Summary

When a guardrail returns a Rewrite outcome for a streaming response, the proxy currently rejects the request with an error:

guardrail {name} requested streaming output rewrite, which is not supported yet

This is a placeholder until whole-response streaming rewrite is implemented.

Context

• Introduced in PR feat(guardrail): add whole replay for streaming chat #108 (feat(guardrail): add whole replay for stream request), specifically in apply_stream_output_guardrails in src/proxy/handlers/format_handler.rs.
• The non-streaming (complete-response) path already supports rewrite via apply_output_guardrail_rewrite.

Goal

Implement support for GuardrailOutcome::Rewrite in the streaming output guardrail path, so that guardrail-driven rewrites can be applied to buffered/replayed stream content.

References

• PR: feat(guardrail): add whole replay for streaming chat #108
• Comment: feat(guardrail): add whole replay for streaming chat #108 (comment)

/cc @bzp2010

[coderabbitai]

🔗 Related PRs

#104 - feat(guardrail): integrate into chat handler [merged]

---

📝 Issue Planner

_{Check the box below or use the @coderabbitai plan command to generate an implementation plan and prompts that you can use with your favorite coding assistant.}

• Create Plan

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord!