Add a timeout to DNS seeder requests to avoid hangs

[teor2345]

Is your feature request related to a problem? Please describe.

When zebrad launches, it asks some DNS seeders for a list of initial peers. But if DNS or the seeder is slow or down, zebrad can hang without any output.

We've observed these hangs in our CI, but we don't know if it's just one seeder that's having issues. If it is, this fix could make our CI more reliable.

Describe the solution you'd like

• in parse_peers, execute each to_socket_addrs call in a separate spawn_blocking tokio task, wrap each task in a timeout, and wait on them completing
• log a warning if any of the tasks time out
• add a timeout argument to parse_peers and initial_peers
• in add_initial_peers, log a warning if there are no initial peers
• check if we still need ZEBRAD_SKIP_NETWORK_TESTS on Windows and Linux CI - this fix might help with their unreliable network behaviour

We don't panic on any of these errors, because they are recoverable - zebrad can discover peers from inbound connections, if its address is distributed by a DNS seeder (or manually).

Describe alternatives you've considered

Do nothing - zebrad will hang if there are DNS issues
Panic - these errors are recoverable, and "no initial peers" is a valid config

Additional context

We discovered this issue while troubleshooting bind hangs in #1535

[teor2345]

Tentatively adding this task to Sprint 2, because:

• we see startup hangs in CI on some platforms - this issue will probably fix them (and definitively give us better diagnostics)
• the fix might allow us to run our full CI on more platforms, which is important, because we've discovered some platform-specific hangs recently
• a hang at startup with no output is a very negative user experience

[teor2345]

I increased the estimate here, because we've added some possible acceptance tests changes.