Populate historical_emission automatically when shifting y₀

[OFR-IIASA]

The parameter historical_emission should automatically be tracked with the value from the variable EMISS when shifting the firstmodelyear, as was explained here in issue #314.

A simple test using the Westeros tutorial (https://github.com/iiasa/message_ix/blob/main/tutorial/westeros/westeros_emissions_bounds.ipynb) showed that the parameter is indeed not populated.

This can be replicated by running the above mentioned tutorial. At the end, clone "scen" and shift the firstmodelyear in the process to 710. Retrieving the parameter "historical_emission" returns an empty dataframe, despite original scenario containing entries in the var(""EMISS).

To replicate, open the above tutorial, and before closing the database connection type:

print(scen.var("EMISS"))
scen2 = scen.clone(model, "test_historical_emission", shift_first_model_year=710, keep_solution=False)
print(scen2.par("historical_emission"))

[khaeru]

This is also linked to #254: the functionality to shift the first model year is currently embedded in the Java code of the ixmp JDBCBackend. This is both a "stack level violation" and means that the code is not currently maintained or updated with improvements in this package.

A fix will involve creating utility code (basically a single function) here that shifts the first model year in a message_ix.Scenario; then the Java code can be bypassed/never used. That function should address this bug.

[khaeru]

I've relabeled this. While it is an intuitive and straightforward thing to expect given the brief/vague description of this argument, the docs also don't promise that this should happen, it's not a bug.