Merge dev into master

[dougollerenshaw]

• Bumps visual_behavior_analysis version from 0.12.0 to 0.13.0
• Brings visual_behavior_analysis in line with AllenSDK 2.10.2

Note that AllenSDK 2.10.2 is not backwards compatible with earlier versions (<2.10). All users must update AllenSDK.

@matchings, @yavorska-iryna and @DowntonCrabby can you enumarate other major changes?

• Add dimensionality reduction module to VBA which currently only includes processing and plotting methods for TCA.
• Add use_filtered_events option to stimulus_response_xr, omission_response_xr, and trials_response_xr in response_processing.py
• Adds functions for plotting behavior & physio timeseries for one session / experiment in visualization/ophys/timeseries_plots.py, along with scripts to generate plots on the cluster
• Adds option to remove Ai94 GCaMP6s data from get_filtered_ophys_experiment_table and sets to True by default
• Adds get_filtered_ophys_session_table with similar filtering & reformatting logic as get_filtered_ophys_experiment_table
• Update BehaviorOphysDataset class to eliminate reformatting of attributes that are now correct in AllenSDK >2.10.0 (The main purpose of BehaiorOphysDataset at this point is to remove invalid ROIs from all cell tables and to provide access to the extended_stimulus_presentations table and behavior movie timestamps and face motion PCs)
• loading.get_ophys_dataset() now has options to get a behavior ophys experiment from lims or from NWB files via the SDK. default behavior is to use BehaviorOphysDataset (which inherits BehaviorOphysLimsApi from the SDK) to provide access to extended stim presentations and filter ROIs
• Creates BehaviorDataset class which inherits a behavior session from the SDK lims API and adds access to extended_stimulus_presentations and extended_trials
• Creates loading.get_behavior_dataset() which gets a behavior session from lims or from NWB files via the SDK, or default behavior is to use BehaviorDataset (which inherits BehaviorLimsApi from the SDK) to provide access to extended_stimulus_presentations and extended_trials
• adapts compute_robust_snr_on_dataframe to work with events in addition to dff_traces
• functionality to add hits and lick bouts to licks table
• functionality to add engagement_state to trials table by merging relevant columns from extended_stimulus_presentations_table (which gets engagement_state from behavior model output files)
• add option to provide a specific time window to get_response_xr functions in response_processing.py, rather than using default values
• add functions to use get_response_xr functions with lick data (similar to functions for getting time aligned physio traces, pupil_area and running_speed, which already existed)
• add functions to get lick_triggered_response_df
• include trace_timestamps as column in output of get_mean_df()
• plotting functions for lick triggered averages in visualization.ophys.container_figures.py
• modify QC plotting functions in visualization.qc.experiment_plots.py to get dataset attributes using SDK rather than using unique functions for each attribute (may need to reconsider in the future if SDK lims API doesnt work for new data types, but in the mean time using SDK lims access methods is preferred)
• adapt pupil plots in experiment_plots.py to use timestamps and pupil traces from SDK rather than EyeTrackingData class
• add plot_flashes_on_traces (using timestamps not frame numbers) to visualization.utils.py

FUTURE TO DO:

• add functionality to get columns available in extended_stimulus_presentations and extended_trials into the SDK so that we no longer need wrappers (BehaviorOphysDataset, BehaviorDataset) to get this information

• potentially eliminate loading ophys_experiment_table from cached file

• make engagement state computeable independent of behavior model output files

• generalize functions for get_X_trials/omission/stim_presentations_df to work with any timeseries as input, rather than having unique functions for traces, running, pupil & licks

[dougollerenshaw]

@matchings : visual_behavior/visualization/ophys/container_figures.py seems like a work in progress. I'm fine merging it in if you'd like. I had to define label = None on line 13 just to get it past the linter, so it seems that the code likely wasn't functional.