Fix VertexAILLM

[stellasia]

Description

The main reason for this PR is that, even if Google GenerativeModel accepts a list of tools as parameter, this is not how we should provide multiple tools. We should instead provide a single tool with multiple function declarations. This PR addresses this behavior.

It also addresses a few other problems related to the way GenerativeModel is instantiated. The usage is:

    # config for chat mode (aka invoke)
    generation_config = GenerationConfig(
        temperature=0.0,
        response_mime_type="application/json",
    )
    # config for tool mode (aka invoke_with_tools)
    # optional
    tool_config = ToolConfig(
        function_calling_config=ToolConfig.FunctionCallingConfig(
            mode=ToolConfig.FunctionCallingConfig.Mode.ANY
        ),
    )

    llm = VertexAILLM(
        model_name="gemini-2.0-flash-001",
        generation_config=generation_config,
        tool_config=tool_config,
    )

Then, internally:

• if invoke mode: we drop the tool config (it's not useful)
• if invoke_with_tool mode, we drop the generation config (otherwise the presence of fields like response_mime_type/response_schema raise errors)

Drawback of this approach:

• We can't have a tool_config per call, which means we can't configure the allowed_function_names parameter per call.

We can't move these parameters to the invoke* methods because then they would be "provider-dependent" and LLMInterfaces wouldn't be interchangeable anymore, which defeats the whole point of the interface.

Type of Change

• New feature
• Bug fix
• Breaking change
• Documentation update
• Project configuration change

Complexity

Complexity: ?

How Has This Been Tested?

• Unit tests
• E2E tests
• Manual tests

Checklist

The following requirements should have been met (depending on the changes in the branch):

• Documentation has been updated
• Unit tests have been updated
• E2E tests have been updated
• Examples have been updated
• New files have copyright header
• CLA (https://neo4j.com/developer/cla/) has been signed
• CHANGELOG.md updated if appropriate

[alexthomas93]

Why are we removing model_params here?