initial estimation checks

The first checks in initial_estimation_checks.m are not robust to the case where for example we use mixed frequency data. Assume we have annual data in time interval 1995-2000 and quarterly data in 2001-2015, I would define the annual variable and the quarterly variable as BOTH observed, but with missing values mutually exclusive, i.e. in 1995-2000 Q data will be NaN, while in 2001-2015 the annual data will have been set to NaN. This means it is not sufficient to check the # of observables declared, but we need to check the max number of non missing data values in each time point, and confront this to the number of non zero shocks declared.