create capabilities to read buffered input data

[martynpclark]

There are too many input operations for cases of calibrating lumped summa simulations across multiple catchments. In this case one data point is read at a time for (potentially) thousands of summa instances -- and these data reads all occur simultaneously. This gives far too many IOPS (Input/Output Operations Per Second).

There is a need to add capabilities to read data from the entire forcing file simultaneously -- for 11 years of hourly data in a forcing file, this can reduce IOPS 100k times.