Add torch library bindings (guard main, add torch wrapper functions)

Update README: clarify kernels binding is stub only

Add torch bindings + build variant torch211-cxx11-cu130-x86_64-linux

Add build variant torch211-cxx11-cu130-x86_64-linux for kernels compatibility

CUDA kernel: zaremba-transfer-operator-cuda

Initial upload: torch-compatible CUDA kernel with pybind11 bindings and CPU tests

initial commit