feat: add optional FlashAttention (FA2/FA3) backend

- add FlashAttnBackend (ATTN_BACKEND.FLASH) using flash_attn_func with KV-cache gather + GQA, mirroring TorchNativeBackend
- add flash_attn_available() probe gated on compute capability plus a real-kernel smoke test, cached at first use
- lazy-import flash-attn via importlib so it stays an optional dependency, raising clear errors when unusable
- add 'flash' optional extra (flash-attn>=2.6) and export the new backend

astrai/extension/__init__.py

@@ -19,6 +19,7 @@ from astrai.extension.attention_backend import (
     ATTN_BACKEND,
     AttentionBackend,
     CudaBackend,
+    FlashAttnBackend,
     TorchNativeBackend,
     attention,
     attn_backend,
@@ -38,6 +39,7 @@ __all__ = [
     "AttentionBackend",
     "CudaBackend",
     "TorchNativeBackend",
+    "FlashAttnBackend",
     "TensorLayout",
     "attention",
     "attn_backend",