feat: add optional CUDA kernel system (csrc/) + fused GQA decode attention

Structure:
  csrc/               -- .cu sources + build.py registry
  astrai/extension/   -- compiled .so + __init__.py (import dispatcher)
  setup.py            -- CUDAExtension from csrc/build.py REGISTRY

Control: CSRC_KERNELS=true|false env var at install time.
Fallback: astrai.extension.available dict for runtime detection.

.gitignore

@@ -7,8 +7,12 @@
 # Allow specific file types and root files
 !astrai/**/*.py
 !scripts/**/*.py
-!scripts/**/*.sh
 !tests/**/*.py
+!csrc/**/*.py
+
+!csrc/**/*.cu
+
+!scripts/**/*.sh
 
 # Allow GitHub files
 !/.github/**
@@ -22,4 +26,10 @@
 !/CONTRIBUTING.md
 !/LICENSE
 !/pyproject.toml
-!/README.md
+!/README.md
+# Allow extension modules (only source .py)
+!/astrai/extension/**/*.py
+
+# Allow build files
+!/setup.py
+!/AGENTS.md