MEKE startup from restart

I noticed that some of our runs that start from previous cases are ignoring the MEKE values in the restart file and are instead using the 'local equilibrium balance' to initialize MEKE. For example, case [268](https://github.com/NCAR/cesm_dev/issues/232) starts from a previous B case, but nevertheless the ocn log file says
```
WARNING from PE 0: MEKE_init: Initializing MEKE with a local equilibrium balance.
```
and the diagnostics show an initial spike in MEKE associated with this initialization

The same thing happens in runs started from a G case, e.g. [274](https://github.com/NCAR/cesm_dev/issues/239).