feat: add BatchNormalization layer to TensorMap model builder

[Sickwoman]

Summary

Adds full-stack support for the BatchNormalization layer in TensorMap,
allowing users to add batch normalization between layers in the model builder canvas.

Changes

Backend

• app/services/model_generation.py — Added custombatchnorm node handler:

elif node_type == "custombatchnorm":
    return tf.keras.layers.BatchNormalization(
        momentum=float(params.get("momentum", 0.99)),
        epsilon=float(params.get("epsilon", 0.001)),
        name=name,
    )(input_tensor)

• tests/test_model_generation.py — Added unit tests for BatchNormalization layer

Frontend

• Canvas.jsx — Registered custombatchnorm node type with default params { momentum: "0.99", epsilon: "0.001" }
• Helpers.jsx — Added validation for custombatchnorm node (momentum and epsilon required)
• Sidebar.jsx — Added BatchNorm drag-and-drop item to the layer sidebar
• BatchNormalizationNode/ — New custom node component with momentum and epsilon input fields

Parameters

Parameter	Default	Description
momentum	0.99	Momentum for the moving average
epsilon	0.001	Small float added to variance to avoid division by zero

How to Test

1. Open the model builder canvas
2. Drag BatchNorm from the sidebar into the canvas
3. Connect it between two layers (e.g. Dense → BatchNorm → Dense)
4. Set momentum and epsilon values
5. Validate and train the model

No Breaking Changes

All existing layer types and model builder functionality are unchanged.

[Sickwoman]

Reopened PR#252 as suggested.