Incorporate median filtered max projection into productionized ophys_etl pipelines (or not)

[danielsf]

Now that #359 has been merged, we have python code capable of generating the legacy max projection images that Leonard and John's code used to generate. We should decide whether or not to make this a part of our regular pipeline. Options include

• Having suite2p_registration automatically call median_filtered_max_projection as a sub-module
• Write a separate queue to run median_filtered_max_projection after motion correction is completed.

We also need to decide if the outputs from this module are going to replace the current, naive np.max(axis=0) maximum projection image being produced by suite2p_registration, or if we want to produce two maximum projection images, a "raw" one and a "filtered" one.