[Mayuri (mayuriphad)]

Fixes #14265

Google Vertex AI recently launched newer Gemini embedding models (such as gemini-embedding-2) that drop support for the legacy :predict endpoint (returning 400 FAILED_PRECONDITION). Instead, they require the :embedContent endpoint using the newer Gemini schema.

This PR updates the VertexAIEmbeddingClient to:

1. Identify if the model uses the Gemini backend (modelId.StartsWith("gemini-")).
2. Conditionally switch the endpoint to :embedContent instead of :predict for Gemini models.
3. Serialize the request using the Gemini REST schema (GoogleAIEmbeddingRequest.RequestEmbeddingContent) rather than the legacy Vertex instances format.
4. Issue sequential requests for batches because Vertex AI's :embedContent only embeds a single text input and does not support :batchEmbedContents.

For backward compatibility, all non-gemini embedding models on Vertex AI will continue to use :predict with batch instances.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[Mayuri (mayuriphad)]

Hi maintainers, a gentle ping on this PR whenever you have a moment to review. I'm happy to make any changes or add tests if something needs adjusting. Thank you for your time!