fix: reconnect when CAS released under KEEP_CONNECTION=AUTO

[paikend]

Summary

• Add checkReconnect() that inspects CAS_INFO[0] status before each request and transparently reconnects when the broker has released the CAS process (KEEP_CONNECTION=AUTO, the default)
• Call checkReconnect() from sendAndRecv() so reconnection is invisible to callers
• Add no-op socket error handler after connect to prevent uncaught EPIPE/ECONNRESET crashes

Root Cause

The CUBRID broker may release the CAS process after a transaction commit when KEEP_CONNECTION=AUTO. This sets CAS_INFO[0] to INACTIVE (0). Subsequent requests on the dead socket cause EPIPE / ECONNRESET.

Approach

Ported the same reconnection pattern used by the official CUBRID JDBC driver (UClientSideConnection.checkReconnect()). The Python driver (pycubrid) was fixed with the same pattern in cubrid-lab/pycubrid#24.

Test plan

• 01_connect — basic connection works
• 02_crud — DDL → commit → DML sequence (the failing scenario) now passes
• 03_transactions — commit/rollback cycles work

Closes #15

🤖 Generated with Claude Code

[paikend]

I just reproduced the EPIPE bug on the current main branch (Node.js v25.2.1, CUBRID 11.2 Docker):

1. Executing query...
2. Query result: 2
3. Closing client...
Error: This socket has been ended by the other party
    at Socket.writeAfterFIN [as write] (node:net:575:14)
    code: 'EPIPE'

For reference, the same pattern was already fixed and merged in pycubrid (cubrid-lab/pycubrid#24), and I verified it works correctly there — both basic query+close and DDL scenarios pass without errors.

This PR applies the same reconnection approach. Would be great to get this merged so #15 can be closed.