Multiome: validation error messages are "lost in all the output"

[brianraymor]

Context

Reported by @brian-mott on sci-data-eng

• Stdout/logging/errors is very busy. I see that the dask cluster has the parameter to only report at the error level, but I will get info messages as well. The actual validation errors get lost in all the output

• For the final error messages, it would be very helpful to have them printed at the end of all the logging/message output, one at a time. Currently, there is the logging Error message followed by another error message with just the list of errors:

WARNING:distributed.shuffle._scheduler_plugin:Shuffle 11e5ccfaf3e21e6743b6ff97807ed253 initialized by task ('shuffle-transfer-11e5ccfaf3e21e6743b6ff97807ed253', 20) executed on worker tcp://127.0.0.1:46761
WARNING:distributed.shuffle._scheduler_plugin:Shuffle 11e5ccfaf3e21e6743b6ff97807ed253 deactivated due to stimulus 'task-finished-1741116843.6624162'
ERROR:cellxgene_schema.atac_seq:Errors found in Fragment and/or Anndata file
ERROR:cellxgene_schema.atac_seq:[None, None, None, None, None, None, 'Anndata.obs.is_primary_data must all be True.', 'Fragment file has duplicate rows.']
INFO:distributed.scheduler:Remove client Client-9d954f19-f92f-11ef-b124-060821df02b7
INFO:distributed.core:Received 'close-stream' from tcp://127.0.0.1:44572; closing.
INFO:distributed.scheduler:Remove client Client-9d954f19-f92f-11ef-b124-060821df02b7