VertexAI Generator: Custom Parameters

[quasi-deus]

Problem

I want to provide my own set of fine tuning parameters to VertexAI generators. This includes tool_config, system_instruction(given that Gemini does not support system ChatMessages) during intialization, and stream(to _send_message) when calling the model.

Solution

To add all these parameters to the function declaration in VertexAI/Gemini generators like for OpenAI generators.

I would be more than happy to create a PR for the same, if this wasn't an intentional design choice.