feat: add Nemotron video-to-text capability

[caihui-lol]

What this PR does

Adds the VIDEO_TEXT_TO_TEXT model capability.
Includes it in the Kubernetes API validation enum.
Maps it between model configuration and OME API capabilities.
Adds it to the NemotronH Nano capability classification rule.
Updates classification and capability-mapping tests.

Why we need it

The imported model nvidia/Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16 supports video-and-text input with text output, but OME currently omits VIDEO_TEXT_TO_TEXT from its detected capabilities.
This causes the imported model metadata to represent only its image, text, and audio capabilities.

How to test

• go test
• E2E test by importing the omni model

Checklist

• Tests added/updated (if applicable)
• Docs updated (if applicable)
• make test passes locally

[slin1237]

Did u run make generate and make manifest?
I think it will update them too?

[caihui-lol]

Did u run make generate and make manifest? I think it will update them too?

thanks @slin1237 for reviewing. I ran both make generate and make manifests. make manifests produced no tracked changes. make generate produced no capability-related output - only unrelated checksum and whitespace-only generator noise, so I excluded them from this PR for review. The current generated schema represents modelCapabilities.items as an unconstrained string, so this enum addition does not change the CRD or Helm manifests.

[pallasathena92]

thanks for contributing, please resolve the conflict and request review again.

[caihui-lol]

thanks @pallasathena92 ! rebase with PR#672 to resolve the conflicts. as PR#672 added a more granular capability AUDIO_TEXT_TO_TEXT to distinguish with AUDIO_TO_TEXT, also update the nemotron rule to use AUDIO_TEXT_TO_TEXT instead. requesting review again.

[beiguo218]

We decided to make the current changes as simple as possible:

1. Just in nemotronHNanoRule, instead of directly adding CapabilityAudioTextToText and CapabilityVideoTextToText, we do a further check for the model type - only add those 2 capabilities when it comes to nemotronh nano omni model;
2. Keep the current rule order and omniRule unchanged.

[caihui-lol]

update nemotron rule to append audioTextToText/videoTextToText when it matches nemotron_nano_omni, requesting review again.

[caihui-lol]

Hi @slin1237 can i get one more review on this PR, thanks so much!