Add MiniMax image-to-video generation

[octo-patch]

Reason: Add the missing MiniMax image-to-video workflow to Media Center.

• Route MiniMax H3 and Hailuo image-guided requests through the matching global or China API version.
• Build first-frame request payloads and parse both v1 and v2 task responses.
• Validate required image input, model selection, regional endpoints, and provider errors before returning results.

Checks:

• pnpm exec bun test apps/server/src/extensions/media-center.test.ts
• pnpm --filter ipollowork-server typecheck
• node .codex/skills/ipollowork-maintainable-code/scripts/audit-changes.mjs
• git diff --check

Experience proof: Incomplete. A live MiniMax credential and approved fraimz voiceover were not available in this worker, so no live provider request or fraimz artifact was produced. To reproduce, configure MINIMAX_API_KEY, invoke video_generate with provider: "minimax", a prompt, and an imageUrl, then query the returned task with task_get.