dttools: use a hash table to map data to array index in priority queue

[JinZhou5042]

In the current implementation of priority queue, the way to get the array index of a data is to traverse the whole array until a match is found, this could be significantly sped up if there is a hash_table in the struct priority to map the data directly to its corresponding index.

[dthain]

No more modifying the priority queue for a while :)

As I said from the beginning:

• First, determine all the actions the scheduler needs to perform on the set of ready tasks.
• Then, design a data structure that satisfies those actions efficiently.

[JinZhou5042]

I agree with you. Just a note, if #4056 wants to have a priority-based temp file replication, for example, files with less replicas are prioritized to replicate, then this might be needed :)