Add expose inputs / outputs feature

[greschd]

Fixes #660.

Adds the functions expose_inputs and expose_outputs to the ProcessSpec. Their options are:

• namespace: expose to a specific PortNamespace, meaning that the ports will be created in that namespace
• exclude: A list of keys which are not exposed
• include: An explicit list of keys which are exposed

The information about which inputs / outputs were exposed is stored in the _exposed_inputs / _exposed_outputs attribute of the spec.

Adds exposed_inputs and exposed_outputs to the Process class. They build up the input / output dictionary corresponding to a specific workchain class (and instance in the case of outputs). Options:

• namespace: The namespace where it is looking for exposed keys.
• agglomerate: Defines whether parent namespaces are also searched.

Finally, it adds an out_many method, which can be used with a dict as returned by exposed_outputs to add many outputs simultaneously.

For additional details see the documentation included in this PR.

[greschd]

Needs https://github.com/muhrin/plumpy/pull/33

[sphuber]

I merged the PR in plumpy. Please rebase on the latest workflows branch to update your branch and up the commit requirement for plumpy in requirements.txt. I will then check it out and run some tests. Thanks

[sphuber]

I think I know why the failing test is occurring. I am putting in a PR now and after that is merged we can merge workflows here again and it should be working

[sphuber]

I have given it a little go on my workchains and for a sub set of the functionality that I need it seems to work. Since the validation is done for the entire input dictionary upon launching, I cannot use agglomerate yet as I would like, but I am not sure if and how that would be possible. I am approving this because what this does make possible seems to be working.