Error instead of panic on timestamp overflow

[progval]

I would have prefered either returning a null value or trying to encode as a non-nanosecond timestamp (sometimes possible if the ORC data does not use nanosecond-level precision), but this would require more significant changes to the code.

[waynexia]

Great 👍 I'm going to merge this

[Jefffrey]

Thanks for this :+1: