Drop get_device_warnings. The DSL noise margin and the firmware on offer
are already facts in get_system_config (dsl_downstream_noise_margin,
firmware_update_available, firmware_version); netOrk derives
dsl_snr_low and firmware_update_available from them (netOrk #1111).
Closes netork#116.
`get_port_forwards` reports IPv4 mappings and IPv6 pinholes together, which is
right — both are inbound rules someone configured. `get_nat_translations` then
iterated all of them, and that is not: a pinhole is a firewall hole, and IPv6
does not do NAT at all.
The entries it produced were nonsense in two ways. They paired an IPv6 host with
the IPv4 WAN address, and `inside_local` came out as "2001:db8::10:22", where
the last colon is a port separator and everything before it is an address that
also contains colons.
The two failing tests were pulling in opposite directions.
`TestGetNatTranslations` was right and the implementation had drifted past it
when pinhole support landed. `TestGetPortForwards::test_returns_all_rules` was
the opposite: it still expected one rule from before pinholes existed, which put
it in direct contradiction with TestIPv6Pinholes further down the same file.
get_device_warnings() now returns only {code, meta} — severity, title,
message, and action are resolved centrally by netork's
WARNING_CATALOG (netork/core/device_warnings.py), not by the driver.
Also renames meta.version -> meta.latest_version for
firmware_update_available so it matches the netgear driver's shape and
the shared catalog message template.
Christian Manivong