Avoid computing annual aggregation results when output not asked

[guilpier-code]

For some studies, It takes a few seconds (between 3 and 10 seconds, more often 4) per year to build annual results by aggregation regarding the binding constraints results :

• when the study contains many BCs .
• when user deactivate the output of marginal costs on BCs

So this overhead is not a big deal. However, we should still consider not computing these results at any time intervals when marginal costs on BCs is not asked by user.

This works a well for any other output variables.