Performance regression in past year with new `promise?` implementation

[bsless]

The new implementation for promise? uses satisfies? and is called on every unwrap.
In high throughput where lots of promises are created and chained, calls to satisfies? come to dominate the code's run time.
We had to either pin promesa versions or drop it at work due to this change.
It seems like an alternative solution could be making unwrap a protocol.
What do you think?