[Feature request] Split node eligibility by scheduler

As of `0.10.4` there is an option to `nomad node drain -ignore-system` and leave the system jobs running.
In our use case when we roll out a new version of AMI (nomad upgrade or any other update), we mark old instances as ineligible, that leads to failure of deploying the system jobs across the fleet if some of our monitoring tools need to be updated (they are running as system jobs).

We'd like to propose an enhancement when eligibility will affect only service/periodic/batch schedulers, but system scheduler will remain fully functional to keep these jobs updatable.