Plotting functionality should join historical_activity to ACT

[glatterf42]

@glatterf42 and @khaeru, I needed to filter the years explicitly here, because the plotting module does not show the values in historical years, which in this case is 690. Ideally, the plot should join historical_activity (for year 690) to ACT (for years 700, 710, and 720). If not possible, the plotting should not print historical year (690) without showing any value for that, which is the default behavior now.

Originally posted by @behnam-zakeri in #815 (comment)

This was said with respect to this line.

[khaeru]

I think we could actually make several such enhancements. For example, the default reporting config includes:

message_ix/message_ix/report/__init__.py

Line 64 in 7dce461

("emi", "mul", "emission_factor", "ACT"),

…but there is also a parameter:

message_ix/message_ix/models.py

Line 433 in 7dce461

par("historical_emission", "n type_emission type_tec type_year")

We can try to do concatenation wherever there are such pairs.

I say ‘try’ because some issues may come up:

• In some cases the dimensionality of the historical_* parameter is mismatched with the product of 1+ variables and 0+ parameters to which it supposedly corresponds. This is true in the case above.
• message_ix_models.report.legacy has a parameter merge_hist, which triggers concatenation of IAMC-structured time series data for pre-model periods with computed values for model periods. This points us to the possibility that a user populates both these historical time series data and the historical_* parameters with potentially different values. What should the reporting do in such cases?