NIST tables should be QTables

As noted in #2918, the current return wavelength unit from NIST depends on the input wavelength. This should be documented better and/or, preferably, should be accommodated through returning a QTable with the appropriate units instead of the current unitless table.

This is a relatively easy fix but should be accompanied by appropriate tests.

Start by locating the NIST query implementation and its existing tests, then review the behavior described in #2918. Done means NIST results return a QTable with wavelength-dependent units, with appropriate tests covering the unit handling.

Written by the indexing model from the issue text.