[ENH] in ModelArray.gam(): check if interaction term with more than 2 levels will display well

[zhao-cy]

Bart, 2022.5.16, slack fixel_analysis: factor smooth interaction: if the factor has more than two levels, there will be more than one interaction term. Couldn’t tell if that will be captured in your naming convention s_x_BYorderedFactor if there were three levels, and level A was the reference level, then gam returns s(x):levelB and s(x):levelC. Each term reflects the difference in the smooth function between that level of the factor and the reference level of the factor. So s(x):levelB is the difference between the smooths for level B and A

[zhao-cy]

resolved by branch enh/test_expected_values