session-wise preproc gets repeated when each session has multiple experiments

[mwaskom]

This is not ideal. In the end it's not a big cost, but estimating the distortions from the fieldmaps is the slowest thing in preprocessing so re-working this would save some computation time. I'm pretty sure it's deterministic but haven't confirmed — if not, it's a bigger problem. But improving this would involve a fairly large reorganization of how things work, so we might just have to live with it.