feat(python): voice-agents v1 — guardrails

[github-actions]

Adds the guardrails recipe for voice-agents v1 (Python).

Demonstrates topic enforcement and response validation guardrails for a voice
agent using prompt constraints and function calling. The agent is restricted to
a support-only scope with two guardrail functions:

• validate_response — checks draft replies for PII before speaking
• redirect_conversation — steers off-topic requests back on scope

Closes part of #565

[claude]

@deepgram/devrel — VP escalation: PR #566 has been stuck for >5 days. Workflow dispatch is blocked (GITHUB_TOKEN 403). Manual intervention needed to trigger tests or merge.

State: labels=[type:samples, language:python], checks=[none]

[claude]

@deepgram/devrel — VP escalation: PR #566 has been stuck since 2026-05-25 with no CI checks running. Re-trigger via workflow_dispatch returned 403. Manual review needed.

State: labels=[type:samples,language:python], checks=none

[claude]

@deepgram/devrel — VP escalation: PR #566 has been stuck for >24 hours and workflow re-trigger attempts failed (HTTP 403 — GITHUB_TOKEN lacks workflow dispatch permission). Manual review needed.

State: No CI checks have run. Needs manual trigger of Lead — Test workflow or direct review.