Allow for constraints to take into account job allocations from another job

### Proposal
It would be nice to have the ability to have "dynamic" constraints on jobs - I guess it would be dynamic, can't really come up with a better term. Mostly just that a constraint should exist that can take into account the existance (or absence) of a running/healthy allocation for another job.

### Use-cases

We run a bit of software that we offer as SaaS and for scaling's sake, it requires some databases to run - at the moment we solve this by having the database (hsqldb) run as a sidecar to the actual job, but this gives us some overhead, and for data synchronisation it requires a bit of a jump-through-hoops approach that, while it works, is ugly as sin.

The easiest solution in our eyes is to run the hsqldb as a system job so each physical node runs one, so the whole pipeline becomes much easier. However, we can't run our SaaS containers if, for some reason, a node does not have the hsqldb allocation running (and system jobs do not seem to be evaluated automatically on a regular basis). It would be great to be able to say in a constraint that a job can only be placed on a node where another job has an active (and healthy) allocation.

No files, tests, or entry points are named in the issue. Start by locating Nomad's constraint evaluation and system-job allocation logic, then clarify how a healthy allocation from another job should affect placement; done requires an agreed design, implementation, and coverage for allocation-present and allocation-absent cases.

Written by the indexing model from the issue text.