when running benchmarks the function map_result make some invalid assumption about the input. (or the caller doesn't correctly order the input).
actually it happens in this PR paritytech/polkadot#3511 that benchmark result are ordered like this:

pallet_collective, council_instance, first_method,
pallet_collective, technical_committee, first_method,
pallet_collective, council_instance, second_method,
pallet_collective, technical_committee, second_method,
....

So the map_result will remove all the methods except the last one.

This refactor make the map_result function more general and simpler. And this should fix the issue with running the benchmark for multiple pallets.

Needed for paritytech/polkadot#3511