backport ExecutionPath.cacheKeyReversed

[yanns]

From #775
and #774

By using this in sangria-slow-logs, we will be able to generate much less objects.

[yanns]

The code I'd like to optimize is the following one (from sangria-slow-logs):

    val path = even(ctx.path.cacheKey)

    queryVal.update(path, ctx.parentType.name, ctx.field.name, success, fieldVal, System.nanoTime())
  }

  def even[T](v: Vector[T]): Vector[T] = v.iterator.zipWithIndex
    .filter(_._2 % 2 == 0)
    .map(_._1)
    .toVector

This new "API" does not really help here.

[yanns]

Actually the reverse iterator can be helpful:

    val path = even(ctx.path.cacheKey.reverseIterator)

  private def even[T](v: Iterator[T]): Vector[T] = {
    val l = new _root_.scala.collection.mutable.ListBuffer[T]()
    var even = false
    while (v.hasNext) {
      val e = v.next()
      if (even) {
        l.prepend(e)
      }
      even = !even
    }
    l.toVector
  }