minor fixes to g.part3 and g.part4

[vincentvanhees]

• GGIR part3 assumes that part 1 output is also available for part 2. I am now revising this such that it uses part 2 as it's reference point.
• GGIR part 4: When now sleep is detected in a day, line 556 is confused, because object spocum.t becomes a character rather than a matrix. I am now revising this line such that characters are converted into matrix.
• GGIR par 5: on a rare occasion struggles to handle missing file in part 4 that is present in part 3

Each seems to affect only a few out of the 4000 files I am currently working with.