Automatic region tracking for ValueObservation

[groue]

This pull request introduces a new very simple way to define a ValueObservation 😍

// NEW!
let observation = ValueObservation.tracking { db in
    // Just fetch and return the values you want to track
}

For example:

// Observe all players
let observation = ValueObservation.tracking { db in
    try Player.fetchAll(db)
}

observation.start(
    in: dbQueue,
    onError: { error in ... }
    onChange: { (players: [Player]) in
        print("Fresh players: \(players).")
    })

The old method tracking(_:fetch:) is still present, but is now presented as an optimized version which can help reduce write contention when you use a DatabasePool and the fetch is slow:

// Optimized for DatabasePool and slow fetches
let observation = ValueObservation.tracking(region, ...) { db in
    // Fetch and return values
}

---

There are two new internal features behind this pull request:

• Observations are now able to automatically infer their tracked region from the executed fetch request(s):

  // Observes the row of id 1 in the player table
  let observation = ValueObservation.tracking { db in
      try Player.fetchOne(db, key: 1)
  }

• The database region tracked by an observation is able to change as time passes and observed values are refreshed.

  This allows to perform conditionals (if, guard, switch) during the fetch.

  For example, the observation below tracks only the team database table, or both the team and player tables, depending on the presence or absence of the tracked team as time passes and database gets modified:

  let teamId = 123
  let observation = ValueObservation.tracking { db -> (Team, [Player])? in
      guard let team = try Team.fetchOne(db, key: teamId) else {
          return nil
      }
      let players = Player.filter(teamId: teamId).fetchAll(db)
      return (team, players)
  }