Add time uncertainty

[GalenReich]

Currently, the tool assumes that the time is known exactly from perfectly accurate image metadata. This may be a reasonable assumption for imagery taken on internet-connected mobile phones with the metadata preserved, but there may be situations where only an approximate time is known.

It would be good if this uncertainty could be represented, perhaps with a broader band.

An outstanding question is how to do this in a computationally efficient way? We should avoid recomputing shadow lengths across the entire globe for different moments in time, if it can be avoided...

[olmonotarianni]

there is a differentiable function somewhere, i guess, to properly compute dt.
or, what about repeating the computation just twice for t = t_0 +- dt, assuming that somewhere in the middle the error was zero (pic)