Raise an error if the deployment strategy changes between runs

It's currently possible to change the deployment strategy in between runs of op-deployer. This doesn't work: in both cases, the pipeline steps will depend on state that doesn't exist. To prevent weird error messages from tripping up our users, we need to check for changes in the deployment strategy and raise a clear error early.

Start by locating op-deployer's run and state-handling entry points, then trace where the deployment strategy is read and persisted between runs. Compare the current strategy with the stored strategy and confirm that a changed strategy produces a clear early error rather than allowing later pipeline failures.

Written by the indexing model from the issue text.