fix(pg): settle query_timeout once via handleError and clear timer

[mikemasam]

Problem
With query_timeout = 100:

// *** callback API — called twice
client.query('SELECT pg_sleep(5)', (err) => {
  console.log(err.message) // 'Query read timeout' x2
})

//*** event API — never called
client.query('SELECT pg_sleep(5)')
  .on('error', (e) => console.log(e.message)) // never fires

Caused by this in client.js:

queryCallback(error)       // direct call
query.handleError(error)   // second call

query.callback = wrapper   // event query now looks like
                           // a callback query, so emit is skipped

Fix

• Timeout settles once through handleError, which picks callback or emit.
• Query emits a new internal cleanup event on success/error to clear the timer.
• Callback is blanked after, so a late database answer is dropped.

// callback API — once
client.query('SELECT pg_sleep(5)', (err) => {
  console.log(err.message) // 'Query read timeout' x1
})

// event API — once
client.query('SELECT pg_sleep(5)')
  .on('error', (e) => console.log(e.message)) // 'Query read timeout' x1