refactor: introduce dedicated refresh queue

[polok]

Background of that change:

• Race condition in the timer setup — guard timer == nil is checked synchronously on the calling thread, while self?.timer = Timer... happens asynchronously on another thread. Two rapid calls to startRefreshing() can therefore both pass the guard and create two timers.
• RunLoop.current.run() on DispatchQueue.global() — this blocks a thread from GCD’s shared global pool indefinitely. With repeated start/stop cycles, this can continuously consume threads from the pool, potentially leading to thread explosion and eventually a crash (EXC_RESOURCE / thread limit exceeded).
• stopRefreshing() does not stop the run loop — invalidate() only invalidates the timer; it does not terminate RunLoop.current.run(). The run loop therefore continues running indefinitely even after the refresh is stopped. This leaves the thread blocked forever and results in a resource leak.