fix(vertex_ai): accept encoding_format param for Gemini/Vertex AI embeddings.

[rodolfo-nobrega]

The OpenAI Python client (v1.x) automatically sends encoding_format: base64 in embedding requests. This was causing UnsupportedParamsError for Gemini embeddings because encoding_format was not in the list of supported params.

This fix adds encoding_format to the supported params list for Vertex AI embeddings. The parameter is accepted but not passed to the Gemini API since Gemini doesn't support it.

Fixes the issue where users get:
litellm.UnsupportedParamsError: gemini does not support parameters: {'encoding_format': 'base64'}

Affected models:

• gemini/text-embedding-004
• gemini/gemini-embedding-001
• vertex_ai/textembedding-gecko
• vertex_ai/text-embedding-005
• All other Vertex AI embedding models

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}