[https://nvbugs/5717993][fix] Add execution_stream across PyExecutor, KVCacheManager, PeftCacheManager to ensure proper CUDA stream synchronization between KV cache transfer operations and model forward kernels. #96289
| Job | Run time |
|---|---|
| 0s | |
| 0s |
| Job | Run time |
|---|---|
| 0s | |
| 0s |