Validation server

[BoySanic]

No description provided.

[DutChen18]

All the validation routes should be examined for race conditions and extra error conditions.

We should also think about how the validator communicates invalid results. Maybe we require the validator to process results in order of submission (this seems like a good requirement anyways to get a consistent canonical result), and say that all results before submitted before the last valid result are invalid, and all results submitted after are not needed.

[DutChen18]

A few more things:

1. There is still a race condition in validate_submit that will take more than just a transaction to solve.
2. I've been thinking a bit about how validators would solve the challenge of generating unique numbers for match ids. It's not that hard, they could just pick increasing numbers. But if we ever decide to store the match ids in the database, that will be a breaking change for validators because they now also have to worry about picking match ids that are consistent with those already in the db.
   Maybe there's a better way: What if the match ids are the id of the assignment (or maybe result) with the first submitted result in that group. This gives groups a canonical match id that will always be the same even if recomputed. It also gives a fancy way of querying for "canonical assignments": SELECT * FROM assignments WHERE match_id = id, even in invalid groups.
3. We need to consider what assignments the validator must resend when a task is run through the validator for a second time. Does it only send assignments that were never sent before? Does it resend all Submitted and Inconclusive assignments? What about if the task is run through the validator again when it already has a canonical result id, does the validator need to resend Valid and Invalid assignments as well? Right now it would, because the match ids are not stored in the database. With my suggestion for generating match ids, it wouldn't, even if the match ids are not stored in the database.

[DutChen18]

There are still some things left to do:

1. Fix the race condition in validate_submit
2. Make validate_submit a privileged route, we can optionally defer this until there is a proper permissions system.
3. Testing.

[DutChen18]

4. Update select.rs with new columns

[DutChen18]

Did 4