[Feature] Immediate data upon signal & wait

[chhwang]

The discussion started from PR #366.

Would it be useful if we piggyback a few bytes (e.g., 4) of immediate data upon a signal()? This could be thought as a similar feature with imm_data in an ibverbs write request. We need discussion on the use cases.

If we need that, while PR #366 implements this by leveraging the semaphore flag, a better way would be just sending additional bytes with the flag. For IB connections, we can leverage imm_data.