Refactor execution input dialog

In prevision for the addition of the workers selection feature, I started refactoring the execution input dialog since it is a rather complex component.

The MR is threefold:

1. Encapsulate things better.

• Move all execution inputs components/models in a dedicated folders (0a7c3ccc)
• Then, move stuff out of the ExecutionInputDialog so that we can better distinguish functional blocks (da513f02, 659917c1, 8f115905)

2. Rework ExecutionInputTableRow type that is at the heart of ExecutionInputDialog. (4c699305)

• The type does not need to be close to the actual API call but must instead only contain fields that are actually relevant to the ExecutionInputDialog features.
• Notably, I added a field target that holds the input target as 'All nodes' | 'All input nodes' | {id: string} to encompass all current usecases. It can easily be extended for future usecases (by label, by task identifier).

3. Add tests for the ExecutionInputDialog. (4eedb931)

• I had to add a bunch of aria-label to be able to find components in the Cypress tests.
• ⚠ This uncovered an important bug that needs to be fixed in a future MR: input of type number are still JSON serialized as string.

Tell me if these refactorings make sense to you. It might easier to review commit by commit.