Add Starcoder2 architecture adapter

[SanjidMzi]

Adds a TransformerBridge adapter for Starcoder2 (Starcoder2ForCausalLM), which had no support (StarCoder1 / gpt_bigcode is already supported).

Starcoder2 is a Llama-shaped decoder (sequential pre-norm, GQA, rotary) that differs from Llama in three ways it shares with GPTBigCode: LayerNorm+bias instead of RMSNorm, a non-gated GELU MLP (c_fc/c_proj), and biases on all projections.

• Adapter + four-place registration (factory, __init__, model registry, report).
• Verified numerically: bridge logits match HuggingFace on tiny-random-Starcoder2ForCausalLM (max abs diff ~1e-7, 100% argmax agreement).
• Unit tests (config, component mapping, GQA weight/bias conversions) + an integration test asserting the forward pass matches HuggingFace.

make format, uv run mypy ., and unit + integration + registry tests pass locally.