Skip short recordings with no sleep estimates in part 5

[vincentvanhees]

When a recording lasts < 2 days with two midnights in it and data_masking_strategy = 2, GGIR part 5 attempts to fix the missing sleep estimates for one night, but encounters a night summary object from part 4 that has only one row with NAs. I am now working on a fix to let part 5 skip the processing of the time series for such recordings.