[Scheduler] Pack `Mutex<TxStatus>` into `AtomicU8`

[hai-rise]

https://github.com/risechain/pevm/blob/131d2745d3305fce8096d1fe52674d62753ab50e/src/scheduler.rs#L42-L44

Rough idea:

• First bit: Being locked or not.
• Next 3 bits: up-to 8 transaction statuses.
• Next 60 bits: up-to 2^60 incarnation index.

This should save 16~18 bytes per transaction and hopefully improve performance.