Line finder improvements

[rwesson]

The line finder is very crude and could be improved:

• a multiple pass approach identifying strong lines first and searching for progressively weaker lines could avoid things like a bright CEL being misidentified as a nearby faint ORL.
• fitting a polynomial function to the shift instead of assuming it's linear might cope better with the uncertainties of wavelength calibrations