feat: reduce stream idle timeout to 1m

[Stebalien]

It's currently 10 minutes. This change will kill all inbound DHT streams that haven't been used in 1 minute.

We keep streams around to avoid the cost of setting up new streams (the multistream overhead, mostly). However, each one of these takes at least one goroutine and some other state.

Ideally, this will let us run multiple DHTs side-by-side without keeping one stream per peer per DHT open.

[vyzo]

SGTM

[raulk]

I’m fine with this as long as we remove old DHT protocol from the mix. That should enable us to pipeline the (sole) protocol re-selection with the request message, thus removing any RTT overhead caused by recreating the stream more frequently. Is that the case?

[Stebalien]

That should enable us to pipeline the (sole) protocol re-selection with the request message, thus removing any RTT overhead caused by recreating the stream more frequently. Is that the case?

Once we've successfully negotiated a protocol once (and after identify), we'll select that protocol immediately without waiting from then on every time we open a new stream. Opening a new stream won't cost us a round trip, it'll just cost us:

1. Some extra bytes for the multistream header.
2. Some extra bytes to open the stream.
3. An extra packet to reset the stream.

[Stebalien]

TL;DR: after identify or successfully creating a single stream, we'll always pipeline the header (https://github.com/libp2p/go-libp2p/blob/76944c4fc848530530f6be36fb22b70431ca506c/p2p/host/basic/basic_host.go#L438)