Fix event loop blocking and deadlocks in batching loop

[tsmith023]

Due to use of threading.Lock.acquire() inside async def __send_batch, the event loop can be blocked from progressing. This becomes problematic when multiple threads call batch.add_object as lock contention can occur between the many locks involved in synchronising the batching algorithm

In certain scenarios this lock contention leads to a deadlock as the event loop becomes blocked by a call to self.__active_requests_lock.acquire() inside async def __send_batch that stops the event loop from progressing thereby causing self.__active_requests to never increase or decrease in value, since all active requests are stuck in a blocked event loop

The same may occur with the other locks so it would be best to migrate all threading.Locks to asyncio.Locks and to call asyncio.Lock.acquire() within the sidecar event loop from synchronous functions

Once this is dealt with, the batching loop is much faster in sending objects. However, this has unintended consequences for the heuristics in the dynamic batch size calculating background thread. The heuristics involved there will need to be refactored to account for the timing behaviour change between using threading.Lock and asyncio.Lock

Draft PR for reference: #1270