Fix Model Selection in batch_predict Endpoint

[sohampirale]

Description:

This PR resolves an issue where the batch_predict endpoint was ignoring
user-selected ML models and always using hardcoded Ridge regression models.

Fixes : #53

Changes Made:

• Created a models_gaze_engineered dictionary for X-axis models that operate
  on the 5-feature engineered input space.
• Reused the existing models dictionary for Y-axis models that operate
  on the 2-feature input space.
• Updated the batch_predict function to correctly pass model parameters
  to predict_new_data_simple.
• Standardized parameter naming to model_name_X and model_name_Y
  for improved consistency and clarity.

Problem Solved:

Previously, users could select different ML models in the frontend, but the
batch_predict endpoint always defaulted to Ridge regression, ignoring
the selected options. This change ensures that the chosen models are
properly propagated through the backend and used for real-time gaze
prediction as intended.

Files Changed:

• /app/routes/session.py — Updated batch_predict function
• /app/services/gaze_tracker.py — Added models_gaze_engineered
  dictionary and updated predict_new_data_simple

Screenshots : All models are working

• Tested this endpoint with all the models we support on frontend(Ruxailab)