feat(agent): add MiniMax text-to-video backend

[octo-patch]

Reason: Raven's video_generate tool lacks a MiniMax text-to-video backend with regional endpoint and async result handling.

Summary

• Add MiniMax V2 text-to-video submission and task polling for MiniMax-H3.
• Add MiniMax V1 text-to-video submission, task polling, file retrieval, and download handling.
• Resolve global and China API endpoints from media or provider configuration.
• Keep download credentials scoped to the configured API origin.

Type

• Fix
• Feature
• Docs
• CI / tooling
• Refactor
• Other

Verification

• Relevant tests pass locally

• Relevant lint / type checks pass locally

• User-facing docs or screenshots are updated when needed

• uv run pytest tests/test_media_gen.py tests/test_tool_registry_timeout.py -x (11 passed)

• uv run pytest tests/test_provider_auth_method.py -x (42 passed)

• uv run ruff check raven/agent/tools/media_gen.py raven/config/schema.py tests/test_media_gen.py (passed)

• uv run ruff format --check raven/agent/tools/media_gen.py raven/config/schema.py tests/test_media_gen.py (passed)

No user-facing documentation update is required because the tool and configuration schema describe the new behavior.

Risk

• Security impact considered
• Backward compatibility considered
• Rollback path is clear for risky changes

Existing video behavior remains unchanged unless a MiniMax text-to-video model is selected. Reverting this commit restores the previous behavior.

Related Issues

N/A