Verify io_uring poller/read/writes and loom pollers

[franz1981]

https://github.com/openjdk/loom/blob/fibers/loom-docs%2Fio_uring.md

Show that both read/write and polling is implemented with io_uring although it seems no batching of submission is happening (e.g. by using the FJ end of batch drain loop), which means that it should pay both the async cost and the syscall one.
SqPoll is usually the solution to it but sadly is both not easy to make it work properly nor so effective as we speak right now.

Having a microbench which shows how the custom scheduler interact with io_uring is a nice have, although not mandatory