[FEATURE REQ] Raise Azure OpenAI's 50-image limit for GPT-6 Astra

### API Spec link
https://github.com/Azure/azure-rest-api-specs/blob/main/specification/ai/data-plane/OpenAI.v1/azure-v1-v1-generated.json

### API Spec version
OpenAI v1 Responses; GPT-6 Astra (`2026-09-03`, GlobalStandard).

### Please describe the feature
Verified with direct REST on 2026-09-06 UTC: `POST /openai/v1/responses` succeeds with 50 inline PNG images, but 51 returns HTTP 400: `Exceeded maximum number of images (50) allowed in the request.` The failing request is only 48,847 bytes (256×128 synthetic images), so this is an image-count limit, not payload size or context exhaustion.

Newer models are incredible at computer use and visual reasoning, and support much longer image contexts. This cap blocks full screenshot and spatial-reasoning trajectories that the models could otherwise process. OpenAI currently documents [up to 1,500 images per request](https://developers.openai.com/api/docs/guides/images-vision).

Please raise Azure's cap to match each model's supported image context and publish model-specific limits. This is a service API feature request, reproduced without an SDK.

Read the linked Azure OpenAI v1 generated API specification first, then compare its Responses image-limit documentation with the reported direct REST behavior for GPT-6 Astra. Done means the service limit is raised to the supported model-specific value and those limits are published in the API specification.

Written by the indexing model from the issue text.