at toplevel: as of now, some commands are evaluated twice,
which conflicts with 'undef'.

executing file.

in doing so, correct two more bugs:
- access to non-existent element of MOM could cause segfault
- mfj took Z from wrong input file, which caused crash when merging just 1 file
merging just 1 file now works, then forbidden

rather, I would need Ruby's each_with_index.

first use of C++ ranged for loop in mem.cpp

  rationale:
  - we accept that Frida will fail for array sizes > 2^15
  - we cannot exclude that complete conversion to size_t would
    cause underflow errors
warning -Wno-sign-compare
replace gsl_sort_index by new function triv::sorted_indices
initialize a few variables to prevent warning

result: no warnings from -pedantic, almost none from -Wall

declaring most indices as size_t
this will be reverted immediately