fix(cuda): add missing imports and required-features test guard for rtx-backend-cuda

P0 fixes:
- Add [[test]] required-features=["cuda"] to Cargo.toml so backend_parity_tests
  only compile when the cuda feature is enabled (avoids E0425 compile errors)
- Add missing Lazy/OnceCell/RwLock/Arc/HashMap imports to device.rs
- Add missing Arc import to tensor.rs
- Add missing Backend/BoolU8/DeviceId/DeviceOps imports to lib.rs

All 16 backend parity tests now pass with --features cuda.

crates/core/rtx-backend-cuda/Cargo.toml

@@ -49,3 +49,7 @@ path = "src/lib.rs"
 
 [lints]
 workspace = true
+
+[[test]]
+name = "backend_parity_tests"
+required-features = ["cuda"]

crates/core/rtx-backend-cuda/src/device.rs

@@ -6,6 +6,18 @@
 use cudarc::cublas::CudaBlas;
 #[cfg(feature = "cuda")]
 use cudarc::driver::{CudaContext, CudaStream};
+#[cfg(feature = "cuda")]
+use crate::CudaBackend;
+#[cfg(feature = "cuda")]
+use rtx_backend::{DeviceId, DeviceOps};
+#[cfg(feature = "cuda")]
+use once_cell::sync::{Lazy, OnceCell};
+#[cfg(feature = "cuda")]
+use parking_lot::RwLock;
+#[cfg(feature = "cuda")]
+use std::collections::HashMap;
+#[cfg(feature = "cuda")]
+use std::sync::Arc;
 
 use crate::{CudaError, CudaResult};
 

crates/core/rtx-backend-cuda/src/lib.rs

@@ -104,6 +104,8 @@ pub use device::CudaDevice;
 pub use error::{CudaError, CudaResult};
 pub use tensor::CudaTensorPrimitive;
 
+#[cfg(feature = "cuda")]
+use rtx_backend::{Backend, BoolU8, DeviceId, DeviceOps};
 use std::fmt::Debug;
 
 /// CUDA backend for RustyTorch++.

crates/core/rtx-backend-cuda/src/tensor.rs

@@ -5,6 +5,8 @@
 use crate::CudaDevice;
 #[cfg(feature = "cuda")]
 use cudarc::driver::CudaSlice;
+#[cfg(feature = "cuda")]
+use std::sync::Arc;
 
 #[cfg(feature = "cuda")]
 /// CUDA tensor primitive - GPU memory storage for tensors.