Force LVT contingency table metrics to use same contingency table structure

[emkemp]

Presently each LVT contingency table metric (PSS, POD, CSI, etc) all store contingency table counts in their own, unique data structure. This wastes both computer memory and code, as brand new data structures and customized read/write subroutines must be written for each metric.

Proposal: Add a general contingency table structure that stores all table counts (hits, misses, false alarms, correct nulls), with support for stratification, and use this structure if any contingency table metric needs to be calculated. Leverage this structure as much as possible (e.g., writing and reading restarts, pulling data from this structure when it is time to calculate a particular metric).

This would be a intermediate term refactoring project (mostly because of conflicting demands on time).

[emkemp]

I still think we should do this. Someday....