transport: Reduce the dialing delay of private addresses

[lexnv]

During the dialing of addresses, the TCP transport does not distinguish between private and public addresses treating them with equal priority.

We have observed in Paseo that validators run with multiple unreachable private addresses. In this case, a validator has been noted for using 30 private unreachable addresses and 1 public reachable address.

The current timeout for addresses is configured to 10 seconds from substrate (as default).

To prioritize public addresses and not waste resources on private ones, I would suggest reducing the timeout to a lower amount, maybe 5 or even 3 seconds for production use-cases. This would still allow sufficient time for connecting on local host; and even on test-nets where network delays and high loads are not reproducible even if 200+ versi nodes.

Part of:

• [ElasticScaling] Enable ElasticScaling on AHKusama and AHPolkadot polkadot-sdk#10425