[Doc] Document num_outputs_per_prompt for /v1/videos

[Hiro208]

num_outputs_per_prompt isn't in any of the video docs, which is what #6899 is
asking about. It's already listed in text_to_image.md and image_to_image.md,
so the omission looks accidental.

It's worth more than just a table row, because the behavior is surprising:
serving_video.py forwards the value, so N videos actually get generated, and
then generate_video_bytes returns videos[0] and logs
"generated %d outputs; returning only the first". Set it to 4 and you pay for
4 videos and receive 1. The note says that plainly so nobody has to find it in
the server log.

Only touching text_to_video.md here, since it carries the canonical
/v1/videos parameter table. image_to_video.md and speech_to_video.md are
shaped differently — happy to add the note there too if you'd rather have it in
all three.

[chatgpt-codex-connector]

Codex usage limits have been reached for code reviews. Please check with the admins of this repo to increase the limits by adding credits.
Credits must be used to enable repository wide code reviews.

[vllm-omni-review-bot]

This PR appears to belong to: docs/design/module/entrypoints.md.

Module owners: @alex-jw-brooks @linyueqian @NickCao

Routing: @alex-jw-brooks via semantic router; @linyueqian via semantic router; @NickCao via semantic router

@Hiro208, please review your own changes and leave a short self-review comment describing what you checked. PRs without author self-review may not be assigned a reviewer.

Please take a look when you have a chance. If you would like an automated review, mention @vllm-omni-review-bot in a comment.

[Hiro208]

Self-review:

• Range and the MiniMax H3 wording match VideoGenerationRequest.num_outputs_per_prompt
  in vllm_omni/entrypoints/openai/protocol/videos.py.
• Traced the value through serving_video.py:347
  (gen_params.num_outputs_per_prompt = request.num_outputs_per_prompt) into
  generate_video_bytes, which returns artifacts.videos[0] and logs
  "generated %d outputs; returning only the first". That discard is what the note
  describes.
• text_to_image.md and image_to_image.md already list this parameter, so the video
  table was the inconsistent one.
• Docs only, no code paths touched. The Read the Docs preview renders the new row and
  the admonition correctly.
• markdownlint clean: MD013 is disabled repo-wide, note body indented 4 per MD007.

[hsliuustc0106]

Verified all claims against main: the value is forwarded (serving_video.py:201), generate_video_bytes logs the discard warning and encodes only videos[0] (serving_video.py:377), and the 1-10 range matches both the protocol Field(ge=1, le=10) and the MiniMax H3 pipeline check. Accurate and well-scoped — thanks for calling out the pay-for-N-get-1 behavior explicitly.