Add pydantic models

[GeigerJ2]

This PR introduces pydantic models for each component of the updated PWD definition, namely PwdInputNode, PwdOutputNode, PwdFunctionNode, PwdEdge, and PwdWorkflow.

It also refactors the AiiDA implementation, adding an AiidaPwdConverter to convert between WG the PWD JSON via the PwdWorkflow intermediate model.

Global inputs and outputs are resolved¹ and null ports in the PWD JSON are handled via an internal __result__ port.

As a side effect of me working on this, I added an additional square step to the arithmetic workflow. This can eventually be moved to a separate PR.

One should note, that in the round-trip conversion PWD -> WG -> PWD, currently, some information still gets lost (see footnote below), while it obviously should not. @superstar54 is working on a PR to make group_inputs functional, which should hopefully fix it (I could also hack my way around here for now, but I don't think that makes too much sense).

Footnotes

1. Well, as far as currently possible, as WG doesn't expose inputs and outputs directly. One can define a mapping of outputs via the group_outputs attribute, see here. An equivalent attribute exists for the inputs, group_inputs, but ↩

[GeigerJ2]

Ping @jan-janssen. Feel free to commit any edits directly here on the PR.

[jan-janssen]

@GeigerJ2 I copied the pydantic model to #97 - it is a simpler implementation which is converting the pydantic model back to a dictionary once it is validated, so the interfaces for the individual codes remains the same.

[jan-janssen]

@GeigerJ2 I am going to close this pull request based on the number of changes in main. Feel free to reopen it once it is ready to be merged.