Calculate historic accuracy

[andrewda]

Ryan suggested being able to see the historic probability of an approach being in actual conditions given a forecast some time out. For example, "This approach was in IMC 80% of the time ceilings were forecast to be 1800 ft at KCVO".

This will assume METAR data is the source of truth, so dependent on #2. We'll also need to store historic forecasts, which we don't currently do.