Consolidate video generation code

[danielsf]

PRs #433 and #435 represent independent code bases for video generation in ophys_etl. This is an unfortunate by-product of the development work we did in summer 2021 (resulting in the framework in #435) and the more staid QC work we've done since autumn 2021 (resulting in the framework in #433). Both frameworks have been tested and validated. #433 produces videos (really, multipage TIFF files) that the scientists want to see. #435 produces the videos that back the cell_labeling_app. Once both have been merged into ophys_etl, we should try to consolidate the frameworks so that they rely on the same backend code. That way, we only have one set of video generation modules to fix if something ever breaks.

Tasks

• Determine how ophys_etl.modules.video code and ophys_etl.utils.video_utils/ophys_etl.utils.thumbnail_video_utils can be consolidated
• Write a single backend to support all instances of video generation in ophys_etl
• As part of the above apply_median_filter_to_video and apply_downsampled_mean_filter_to_video should be moved into a single video_filters module (that name is open for revsion)
• Connect ophys_etl.modules.video and ophys_etl.utils.thumbnail_video_generator to the new backend

Validation

• ophys_etl.modules.video still produces the downsampled multipage TIFF files the scientists want to use to QC motion correction
• ophys_etl.utils.thumbnail_video_generator still produces the thumbnail videos used by the cell_labeling_app