- Add TEST_DATA_SEND_RECV verifier kernel that replays fill_data PRNG
with peer_rank seed to validate received data
- Add double-buffer support for sendrecv in executor_test.py:
allocate 2 input/result/test buffers, alternate per iteration
- Create two executor funcs for sendrecv, one per buffer pair
- Update bench_correctness and bench_time to handle double-buffer
- Add bandwidth reporting to output
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>