Make PART_VARS more consistent (was: integrate participation jump codes in int_unexp_records_segment)

Make the following more consistent:

We have PART_VARS on segment level and on item level metadata, but they work very different.

We have the old idea of hierarchical status variables (PART_VAR on item-level set to 1 -> participates in the segment, 0 -> does not p.). If this PART_VAR has a PART_VAR assigned again, then this makes up a hierarchy, and as soon as the path from the item to the "root" of this hierarchy contains at least once a 0, no observation is expected.

We have also the new idea: on item level, we use missing codes for qualified missingness mapping them to AAPOR codes. Any value not being a missing code means participation. On the segment level, we assign here the other sort of PART_VARS, in this case, this is a variable, which only contains missing codes (plus one special "missing code" that says "not missing"). This variable is then used to compute qualified segment missingness after mapping these missing codes (for the segment_level.SEGMENT_PART_VARS) to AAPOR codes.

Ideas:

• make the "old" PART_VAR entry also use missing codes to be consistent with the other sort of part-vars -- but find a way for possible backwards compatibility (like having two missing codes, 0 = anything not participation and 1 = I.
• maybe add an explicit hierarchy possible for segments, e.g., introducing a new column PARENT_SEGMENT. A conversion from old item-level part-vars hierarchies may be feasible.

old:

The current implementation of int_unexp_records_segment only distinguishes NAs from measurement value in the variables, but the user should be able to specify a list of jump codes that define the participation in each segment