Because:
- We keep seeing random timeout errors in production.
- It's better to retry operations that time out a few times before
  giving up completely - this runs in the background sending cron mail,
  so we should try to take care of our own problemns ;-)

Because:
- simeple_rest_client raises generic "ClientConnectionErrors" on read
  timeouts. The real reason is only discoverable by checking the
  exception's __cause__.

Because:
- It's not clear whether the implicit timeout in the underlying library
  is "infinite" or some arbitrary value, but we keep getting timeouts
  in production.

Because:
- The newest version of simple_rest_client seems to have requirements
  that cannot be satisified.
- We don't know that this code is compatible with simple_rest_client
  1.0.0

Because:

- NAV 5.0 drops SNMP attributes as part of the Netbox model, introducing
  the concept of management profiles. Kindnavsync needs to be able to
  work with both pre-5.0 and post 5.0-versions of NAV for now.

Support for older NAV versions can be dropped later.

Also update to a trove license classifier that matches the actual license.

Viste seg vanskelig å debugge disse feilene mot en testserver :)

also ditch IPy in favor of ipaddress, which is built-in in Python 3

having timestamps in the log output will cause kindnavsyncdiff to trigger on
every run, even though the output is in fact the same.

if devices that exist both in Kind and NAV have differing rooms or equipment
types, update the NAV devices to be in sync with Kind.