refactor: unify kernel module loading and packaging

- loader.py: lazy/cached import; is_available defers the actual load; get_module raises on unavailable
- ops/{attention,rotary,fp8}: use get_module instead of touching private _modules or their own _mod() cache
- package-data: ship astrai.extension.lib *.so in built wheels (non-editable installs previously lost every kernel)

pyproject.toml

@@ -37,6 +37,9 @@ flash = ["flash-attn>=2.6"]
 [tool.setuptools.packages.find]
 where = ["."]
 
+[tool.setuptools.package-data]
+"astrai.extension.lib" = ["*.so"]
+
 [tool.setuptools.dynamic]
 version = { attr = "astrai.__version__" }