Make disk IO and processing concurrent

[ashwin95r]

In processTask fetching the values and processing it were sequential which was slow when we had huge UID lists.

For example, this query on SuperUser data (50M rdfs) has ~350K UIDs which are of the type Question

{
  var(func: eq(Type, "Question")) {
    a as count(Upvote)
    b as count(Downvote)
    sc as math(cond(b > a/2, a, 0))
  }
    
  contr(func: uid(sc), orderdesc: val(sc), first:100) {
    val(a)
    val(b)
    val(sc)
    Title{
      Text
    }
  }
}

The processing time reduces from ~20s to ~9s when we make GetOrCreate and the processing concurrent.

---

This change is 

[ashwin95r]

After testing it out a bit more, just doing the decr() and releasing the PLs at the earliest gives us most of the benefits. So we don't have to use channels to fetch the PLs separately. Continuing this in #1255.