Understand `g`

[MoLi7]

@jvendries suggested to

• Compare g_2017 vs. g_new_2017 = derive_gross_output_after_redefinition(2017), and/or

• Compare g_eff_old_2018, g_eff_old_2019, etc. vs. g_new_2018, g_new_2019, etc.?

Comparison 1 would help us confirm whether the changes we are seeing are from the change in method (which should not be the case, given the output of test_2017_redefinition_roundtrip() ).

Comparison 2 would allow us to see the variability by year between the new and old methods to see which year the values start to diverge significantly.

@MoLi7 suggested to add them in /publish as ad-hoc analysis.