Update evaluators to consider NaN for evaluating upper/lower bounds

[yyanyy]

See discussion and the test case, due to the comparison implementation of ORC stats, for float/double columns in ORC files, if the first value in a file is NaN, metrics of this file will report NaN for both upper and lower bound despite that the column could contain non-NaN data. This means that today we may unexpectedly skip including an ORC file for predicates that utilize bounds. We need to update evaluators to consider this edge case.