Motivation
Based on the current implementation I am working on for #54, I am making the decision to start by only supporting adding new steps in series, not in parallel. Pipeweld, the solver that underlies this logic, already supports adding steps in parallel, we mostly just need to write the integration for this aspect. There's also a similar integration to handle an edge case where we need to move an existing step to break up a parallelism.

Supporting parallelism makes pipelines run more efficiently by taking into account (the lack of) dependency relations between steps. Supporting the ability to break up an existing parallelism allows us to satisfy otherwise-impossible dependency relations.

Summary of desired enhancement
Handle adding a new step to a Bitbucket pipeline in parallel to an existing step, and breaking up an existing parallelism.