PWGHF: Add multiplicity and BDT output scores to Ds task

[fchinu]

This PR adds multiplicity and machine learning information to the output of the Ds task into a THnSparse.

This is done by adding templates to the original functions in the task, which allow to use/don't use ML and multiplicity tables. It is possible to choose which one to use by turning on the relative process function. It is also possible to choose which ML classes to save and which detector to use to evaluate the multiplicity percentiles via configurables.

In order to loop on the selected candidates which may or may not have ML information, several partitions have been defined (as Partition<CandDsMcReco> does not have ML information, but Partition<CandDsMcRecoWithMl> has it, for example)

Also, refactoring of part of the code was made, making it easier to modify and maintain.

[fcatalan92]

Hi @fchinu, thanks a lot for the developments and refactoring! I have a few comments.
The relevant one is that on the process function (and option) names. IMHO, if their pattern is different it is easier to not make mistakes when configuring the task.
Another thing to keep in mind is that, at the moment, the multiplicity information is not filled for the MC matched candidates. This will be needed in the future for analyses vs multiplicity.