tests/socket_zep: fix automatic test

[benpicco]

Contribution description

The automatic test for socket ZEP was broken because it did not run on Murdock. This PR attempts to fix this:

• use the API as expected by the Radio HAL
• supply a EUI-64 so the native instance always uses the same address
• switch to IPv4 as Murdock does not support IPv6. This allows to run the test in CI

Testing procedure

make -C tests/socket_zep test

should now work again.

Issues/PRs references

fixes #16946 (comment)

[kfessel]

Changes seem sound,
Test in CI was run.
https://ci.riot-os.org/RIOT-OS/RIOT/17432/79891e93181dca6fc2f2086dce0e49fec2566499/output/compile/tests/socket_zep/native:gnu.txt

Just a small question

[kfessel]

Why did you remove this part of the test ?

[benpicco]

With the radio HAL sending a NULL buffer will no longer result in any data being transmitted. Previously this would send a header-only packet (and this test would verify that).

[kfessel]

maybe:

Elements of this list may have iolist_t::iol_size == 0 and (in this case only) iolist_t::iol_data == 0

should be tested instead

[benpicco]

I think this makes no sense. We are on layer 2 here, so for a real device this would mean no data on the bus / air.
Only with the (old) ZEP implementation this would write the ZEP header (that wraps the IEEE 802.15.4 frame).

Mind you that we are using the sub-mac now which does make assumptions about the contents being transmitted (namely them being a IEEE 802.15.4 frame) - we'd now need to add checks there, e.g in sys/net/link_layer/ieee802154/submac.c:414 where it checks if the ACK_REQ bit is set - for a case that will never happen in reality.

[kfessel]

changes are sound, test ran by murdock, removal of unneeded test explained -> question resolved