Investigate libp2p swarm halting after many concurrent streams are open in p2p_stream from a single peer.

[CHr15F0x]

This is a follow up issue for #2334.

Whilst limiting the default max number of concurrent streams in p2p_stream for 1 peer to 1 mitigates the issue of swarm halting we are still not sure what the root cause is.
There are 2 suspects here:

• p2p_stream implementation contains a bug,
• libp2p stream is susceptible to being overrun with events when too many concurrent streams are being handled and the queues are too small (by queues I mean these two swarm config settings: notify_handler_buffer_size and per_connection_event_buffer_size); The expectation here is that any limits enforced in the config should result in the other peers' qos being degraded, with "our" libp2p-swarm instance still being responsive and handling events.

Understanding the exact mechanism of the problem is crucial to ensuring DoS resistance in this scenario (many concurrent p2p_stream request/response streams).