Refactoring fill xarray function

[huche6]

The idea here is to drastically improve the efficiency of the function fill_xarray_from_input_parameters
Previously, it iterated through all indexes of an initialized xarray DataArray called array.

What I propose here is to keep the same iterative process but instead of allocating values directly to a DataArray, I append dictionnaries into a list. This list of dictionnary is then used to create a pandas DataFrame.
For each column of interest (powertrain, size, value, year, parameter), I flatten (through explode function) the lists contained within each row of the dataframe. I use a dummy variable "g" which count elements in each lists for each row of the DataFrame.
The final DataFrame is built using the flatten dataframe and the corresponding parameters data. NaN are automatically used when a DataFrame is broadcasted. Then I use the ffill method from DataFrame.fillna to replace the NaN by the expected values.
The columns size, powertrain, parameter, year and value are then passed as index of the DataFrame.
The resulting DataArray is created from the final pandas Series obtained.

It was not easy to detailled the process here, do not hesite to ask further question if needed.

All tests in carculator_utils, carculator and carculator_trucks seem to work properly.

[romainsacchi]

Awesome, thanks @huche6