lower memory footprint for "fill the demSeOth with zeroes so that oneRegi tests work"

[lecfab]

Following @orichters' comment in #637, this PR reduces the memory footprint: v_demSeOth is now filled with zeroes up to a size 19x21x3 (the previous version filled it to 24000 instead of 3).

[fbenke-pik]

Not sure if it is necessary to create a dummy magclass object just for the matching here. Usually, the reference ref passed to matchDim is the magclass object you want to bind v_demSeOth to (i.e. which causes mbind to crash).

[lecfab]

Hey,
I tried to matchDim v_demSeOth with tmp1, but the dimension names don't match. That's why i used an intermediate named object.
What would be a more fitting solution?

[lecfab]

Falk's solution here: #644

[lecfab]

@fbenke-pik For reproducibility, the error happens with the gdx /p/tmp/fabricel/powerRealizations/output/testOneRegi/fulldata.gdx:

Error in mbind(tmp1, setNames(dimSums(mselect(v_demSeOth, all_enty = "seh2"),  : 
  Cannot handle objects! Spatial as well as data dimensions differ!
  Differences from first mbind() input:
  spatial:
      missing: `LAM`, `OAS`, `SSA`, `NEU`, ...
       having: `EUR`
  data:
      missing: `SE (EJ/yr)`, `SE|Biomass (EJ/yr)`, `SE|Electricity (EJ/yr)`, `SE|Electricity|Combined Heat and Power w/o CC (EJ/yr)`, ...
       adding: `SE|Input|Hydrogen|Other Energy System Consumption (EJ/yr)`

Error traceback is

1. convGDX2MIF(gdx)
2. convGDX2MIF.R#53: mbind(output, reportSE(gdx, regionSubsetList, t)[, t, ])
3. reportSE(gdx, regionSubsetList, t)
4. reportSE.R#401: mbind(tmp1, setNames(dimSums(mselect(v_demSeOth, all_enty = "seh2"), dim = 3), "SE|Input|Hydrogen|Other Energy System Consumption (EJ/yr)"))