add a new method iter_process_result which does the same as proces_result but is a generator (closes #1625374) process_result is reimplemented using the new method, and the two helper methods are turned into generators. These generators use cursor.fetchmany instead of cursor.fetchall after setting cursor.arraysize to 100. This means that the whole result set should never loaded in memory when using the iter_process_result method. This is used in the "portable" database dump implementation when we typically to 'SELECT * FROM table', but could probably be used too in other parts of cubicweb.

debian/tmp/usr/bin/cubicweb-ctl usr/bin/
debian/tmp/usr/lib/PY_VERSION/site-packages/cubicweb/cwctl.py usr/lib/PY_VERSION/site-packages/cubicweb
debian/cubicweb-ctl.bash_completion etc/bash_completion.d/cubicweb-ctl