Improving stop-to-shape validation

[emmambd]

In conducting a distribution analysis of stops min distance from shape for https://github.com/MobilityData/gtfs-validator/issues/1613, we noticed that there are feeds that don't return trip_distance_exceeds_shape_distance that seem to have huge discrepancies between the stop and shape, e.g 150 kilometres.

There's an opportunity to go further in our shape validation to flag these cases.