[MRG] feat(model): Update gemini.py to use new Gemini API

[syangx38]

Closes #302

advisor.py
remove the duplicate schema_web_search

gemini.py uses the google-genai SDK, highlights:

1. gemini api migrate doc states the new SDK promotes a stateless approach as its primary interaction model, which is a significant change from the stateful sessions that were central to the old SDK. Specially, instead of model.start_chat, now it's client.models.generate_content()
2. As pointed out in the issue, the new Gemini API function calling requires an explicit process to first get a tool call suggestion and then a separate API call to get the final structured output (AdviseAgent etc all requires structured JSON to work).
3. Instead of using recursive (as openai.py does), used iterative loop to manage the multi-turn conversation for tool calling.
4. def query() method logic:

def query(chat_history, **kwargs):
    // Initial Setup
    system_instruction, prompt
    tools
    json_output_required = False // Flag to force JSON output

    // Tool-Calling Loop
    FOR LOOP:
        config = json_only_config IF json_output_required ELSE base_config
        json_output_required = False

        response = call_gemini_api(prompt, config)
        function_call = get_function_call_from(response)

        IF function_call:
            execute the local function            
            append model's response part to prompt
            append tool_result
            json_output_required = True // Set the flag 
        ELSE:
            // The model returned text
            final_response_content = get_text_from(response)
            BREAK 

    // Return the raw string content for the calling agent to handle
    RETURN final_response_content

---

What has been done to verify that this works as intended?

sample logs below

Applying feature engineering...
Starting training with 5-Fold Stratified Cross-Validation...
--- Fold 1/5 ---
--- Fold 2/5 ---
--- Fold 3/5 ---
--- Fold 4/5 ---
--- Fold 5/5 ---
Overall OOF AUC: 0.88561
Submission file created successfully at ./tabular-playground-may-2022/public/submission.csv
First 5 rows of submission file:
       id    target
0  800000  0.609488
1  800001  0.440478
2  800002  0.892964
3  800003  0.574231
4  800004  0.362269

Why is this the best possible solution? Were any other approaches considered?

Checked as the official doc required https://ai.google.dev/gemini-api/docs/migrate extensively.

How does this change affect users? Describe intentional changes to behavior and behavior that could have accidentally been affected by code changes. In other words, what are the regression risks?

Will implement gemini users, but users can switch to other models.

Do we need any specific form for testing your changes? If so, please attach one.

No.

Does this change require updates to documentation? If so, please file an issue here and include the link below.

No.

Before submitting this PR, please make sure you have:

• confirmed all checks still pass OR confirm CI build passes.
• verified that any code or assets from external sources are properly credited in comments and/or in
  the credit file.

[leeeizhang]

LGTM! many thanks

[YuanmingLeee]

Also fix lint:
./mle/model/gemini.py:116:37: W291 trailing whitespace
./mle/model/gemini.py:129:54: W291 trailing whitespace
./mle/model/gemini.py:156:90: W291 trailing whitespace

[huangyz0918]

Thank you @syangx38 Summer! Nice one 👍