When using multiple instances of the same pallet, each instance was
executed with the components of all instances. While actually each
instance should only be executed with the components generated for the
particular instance. The problem here was that in the runtime only the
pallet-name was used to determine if a certain pallet should be
benchmarked. When using instances, the pallet name is the same for both
of these instances. The solution is to also take the instance name into
account.

The fix requires to change the `Benchmark` runtime api to also take the
`instance`. The node side is written in a backwards compatible way to
also support runtimes which do not yet support the `instance` parameter.

---------

Co-authored-by: GitHub Action <action@github.com>
Co-authored-by: clangenb <37865735+clangenb@users.noreply.github.com>
Co-authored-by: Adrian Catangiu <adrian@parity.io>