Fix idle ping spin loop from exhausted AsyncStream iterator

[datlechin]

Summary

Fixes the root cause of #618 (excessive idle ping traffic, ~30 pings/second instead of 1 per 30s).

Root cause

The health monitor used a task group to race Task.sleep(30s) against AsyncStream.Iterator.next():

await withTaskGroup(of: Bool.self) { group in
    group.addTask { try? await Task.sleep(for: .seconds(30)); return false }
    group.addTask { _ = await wakeIterator.next(); return true }
    if await group.next() != nil { group.cancelAll() }
}

When the 30s sleep won, group.cancelAll() cancelled the wake-up task, causing next() to return nil. An AsyncStream iterator that returns nil is permanently exhausted — all subsequent next() calls return nil instantly. This turned the monitoring loop into a spin loop.

Fix

Removed the AsyncStream + task group race entirely. Replaced with a simple Task.sleep loop:

while !Task.isCancelled {
    try? await Task.sleep(for: .seconds(Self.pingInterval))
    guard !Task.isCancelled else { break }
    await self.performHealthCheck()
}

The checkNow() early wake-up method was dead code (never called anywhere in the codebase), so the AsyncStream mechanism served no purpose.

Also added diagnostic logging to the ping handler for ongoing monitoring.

Closes #618

Test plan

• Connect to PostgreSQL, leave idle for 2+ minutes — verify pings fire every ~30s (check Console.app filter ConnectionHealthMonitor)
• Disconnect network → verify auto-reconnect triggers with exponential backoff
• Reconnect manually via toolbar → verify single monitor running