Module gemma 

Gemma Model V2 - Clean implementation using solid abstractions

This implements the Gemma architecture which is used by 10+ models including:

• Gemma-2B, Gemma-7B, Gemma2-9B, Gemma2-27B, CodeGemma, PaliGemma
• Uses unified Tensor type from tensor_core
• Implements Model trait from model_core

Structs

GemmaAttention

Gemma attention mechanism

GemmaConfig

GemmaLayer

Gemma transformer layer

GemmaMLP

Gemma MLP (feed-forward network)

GemmaModelV2

Main Gemma model implementation