Edge case with wsV2 and high throughput subscriptions

[fselmo]

What was wrong?

• There's an edge case for WebsocketProvider (V2) with high throughput subscriptions (e.g. newPendingTransactions) where if there is even a single call for each transaction, there is enough wait time on calls where the subscription queue will get backed up and eventually the _listen_event will trigger and not work as it should. It basically bricks the listener task and the only current way to bandaid "fix" it is to increase the cache size - though this isn't sustainable because it will eventually fill up to that higher number as well and brick again.

How can it be fixed?

Look into the way the _listen_event (asyncio.Event) is working and figure out a good way for the waiting to work with this scenario of a full subscription queue that keeps getting more full. Maybe wait until the queue is half full to start listening to messages again? An arbitrary wait time like 1 second? We'll have to play around with this and find a good balance.