ACCESS-ESM restart repro tests don't work

[dougiesquire]

The test_restart_repro test compares checksums from a 2x1day run vs 1x2day run. This doesn't work for ACCESS-ESM1.5 or ACCESS-ESM1.6 for two reasons:

1. The CICE cice_in.nml::dumpfreq is not modified by the model test class' set_model_runtime method, so no CICE restart is written after one day. This is easy to fix.
2. The UM can only write restarts at the end of a month (apparently). This is harder to find a solution for. Obviously we could extend the restart repro run to 2x1month vs 1x2month, but then a restart repro test of ACCESS-ESM1.6 takes approximately 50 mins (and fails... but that's a separate issue).

cc @blimlim

[blimlim]

Happy to report that I wasn't correct about the UM restarts. The atmosphere/namelists file has a DUMPFREQim=-9999,0,0,0, setting which controls the restart write frequency. Here -9999 sets it to write at the end of each calendar month. When given a positive variable, it uses units of timesteps.

Setting DUMPFREQim=48,0,0,0, to write daily restarts wrote the following files during a 3 day run:

aiihca.daa1120  aiihca.daa1130  aiihca.daa1140

I the UM side should be easy to fix too!

[dougiesquire]

Here -9999 sets it to write at the end of each calendar month

So intuitive... 😒