check async atomic-based mutexes

[jimy-byerley]

At the moment

• we are using a lot of std::sync::Mutex in RawMaster and some other places, they are not never held long so they should not perturbate the async flow, however it is unclear what their locking/unlocking performance cost is.
• we are using tokio::sync::Mutex in Group and in crates using etherage. they are meant to not perturbate the async flow, but their locking/unlocking cost can be high since they are relying on a std::sync::Mutex that is locked/unlocked at each poll.

to investigate

We should see if it is possible to replace both of them with a home-made async mutex implemented with atomics, so polling it would only be an atomic read