ollama: vision: remove prefix info from base64 data

[jakobdylanc]

When using ollama with the llava vision model, you will get an error when including prefix info in the base64 image data (ex. data:image/jpeg;base64,) per the OpenAI API format. LiteLLM can remove this prefix info and prevent the error.

Solves this issue: https://github.com/BerriAI/litellm/issues/1427