handle constraints from run_exports

[minrk]

• recognize weak_constrains and strong_constrains keys in run_exports
• explicitly handle run_exports lists and dicts separately, rather than conflating unrecognized dict keys with weak list entries. Now unrecognized keys are ignored with a warning rather than interpreted as weak requirements

I initially added the constraints to the dependencies for the solve, since right now they are ignored, but I'm not sure that's the right thing to do, so I backed it out.

previous handling added any unrecognized keys (e.g. strong_constrains) to weak requirements, resulting in:

Could not solve for environment specs
  The following package could not be installed
  └─ strong_constrains does not exist (perhaps a typo or a missing channel).

seen in the petsc321 migrator in dolfinx_mpc, specifically this run.

[beckermr]

Thanks for this! Edit: we need to add the constrains bits as predicates to the solver.

@jaimergp should be able to guide us on how to do that.

[beckermr]

Sorry @minrk. My initial comment was wrong and we should not be adding constraints to the host or run deps. I will back those commits out.

[beckermr]

Back these out.

[minrk]

Makes sense, sorry.

I think I see how to apply constraints as actual constraints instead of dependencies, if you want to go that way. I can do it as another PR, since this fixes the immediate problem and nobody has complained about lacking constraints so far, as far as I know.

[jaimergp]

Sorry, just saw the ping. I guess it's not necessary anymore to discuss the solver constraints?

[beckermr]

We want to inject them into the solver @jaimergp so yeah any help is appreciated.

Let's definitely move to another PR @minrk !

[jaimergp]

Cool, ping me in the new PR once ready :)