rsky-relay: clear events before polling & rsky-relay: don't try to reconnect if failed to requestCrawl

[utkarshgupta137]

Summary

Unlike mio crate, the polling crate doesn't automatically clear the events vec before waiting. This means that we weren't using epoll properly and were always iterating all the connections every time. This also made back-filling unsymmetrical because we were always trying to read from the connections in the same order.
Also, stop trying to reconnect crawler if the error occurred during connection phase. This fixes loops for some hosts that seem to be offline (yet the bsky relay listHosts lists them as active). In the future, a more advanced retry logic could be added to handle these cases once again.
Also tweaked some trace levels where appropriate.

Changes

• Feature implementation
• Bug fix
• Documentation update
• Other (please specify):

Checklist

• I have tested the changes (including writing unit tests).
• I confirm that my implementation aligns with the canonical Typescript implementation and/or atproto spec
• I have updated relevant documentation.
• I have formatted my code correctly
• I have provided examples for how this code works or will be used