fix(connection): retry failed session establishment

[faysou]

Summary

Keep the sync and async reconnect loops active when the socket reopens but session establishment fails. This covers the interval where TWS accepts TCP connections before its API handshake is ready, such as during an automated restart.

stateDiagram-v2
    [*] --> ReconnectSocket
    ReconnectSocket --> EstablishSession: socket connected
    ReconnectSocket --> Backoff: socket error
    EstablishSession --> Connected: session ready
    EstablishSession --> Backoff: handshake or startup error
    Backoff --> ReconnectSocket: attempts remain
    Backoff --> Failed: attempts exhausted

Loading

A reconnect succeeds only after the complete session is ready. Transient establishment failures consume an attempt and follow the existing backoff policy.

Testing

• cargo test -p ibapi --features sync.
• Sync and async tests with an incompatible handshake followed by a valid handshake.
• Paper TWS restart smoke: reconnect survived the startup window and downstream subscriptions resumed.

[wboayue]

Thanks @faysou. Retrying through the TWS startup window where the socket accepts but the handshake isn't ready is right.

I pushed two small additions to your branch:

• reconnect() now returns the last attempt's error on exhaustion instead of a bare Error::ConnectionFailed, so a permanent cause (e.g. incompatible server version) is named rather than hidden. The dispatchers still produce the ConnectionFailed sentinel, so loop-exit behavior is unchanged.
• Merged main and added a CHANGELOG entry under [Unreleased].

Will merge once CI is green.