Reading and validating inputs

[JMMackenzie]

Extend the code in utils.py to include validation for inputs, and potentially handling containers of inputs (mapping a TREC run file to either a list of lists, or a dictionary of lists, etc).

Validation needs to ensure that:

• Document ranks are obeyed -- ranks are assumed to be strictly increasing, but gaps are allowed for representing tied elements. For example, having a sequence of ranks like [1, 1, 1, 4, 5] should be valid,
• Scores are used only as a diagnostic against provided ranks. If the ranks and scores do not agree, either warn or error.

[JMMackenzie]

Input file types need to be coerced into the appropriate type given the metric of choice.

Rankings

• If a qrel file is to be treated as a ranking, we can format it into an RBRanking with the highest grade as the first (tied) group, and so on.
• A trec run is naturally a ranking.

Sets

• A qrel file is naturally a set in some sense, but we will convert to positive/negative sets according to some cutoff.
• A trec run is treated as all positive; things not in the run are negative implicitly.