Initialization weirdness with SolidOxideCell and Sequential Decomposition

[dallan-keylogic]

trafficstars

In #1625, a user reported degrees of freedom changing after initializing his model using the SequentialDecomposition. Upon further investigation, it looks like the oxygen inlet variables are improperly remaining fixed after initialization. I thought it might be a bug with the SOC initialization, but it looks like we already have a test that unfixed variables are reverted unfixed after initialization. Maybe the problem happens only some of the time? Maybe it's the sequential decomposition?