To pass the test on the value of the approximated log-likelihood.

This test should not be included in the testsuite (far too slow)

The mod file is not yet part of the testsuite sicne something is
clearly wrong here.

We verify that, in a linear model, the likelihood value obtained using
the SIS algorithm is equal to the likelihood value calculated with a
standard linear Kalman filter.

The value of the likelihood at the mod will be used to test various
nonlinear filters.

Under oo_.likelihood_at_initial_parameters.

 Fix crash of perfect foresight w/ expectation errors + marginal linearization + several learnt_in blocks

[skip ci]

Ref. !2350

Prevents cryptic crashes

Leads to a simpler code.

I am not sure I want to repair this...

Hybrid stochastic extended path with the constant correction in the
reduced form of the model approximated at order k>1 (perturbation
around the deterministic steady state).

In each iteration using the solution from previous step. Previously
the initial guess for the auxiliary stochastic perfect forsight model
was the steady state in all periods and (future) worlds.

The initial condition of the state variables was not updated in each iteration.

For some reason, the path expected by the -B option of gfortran should now
be in Windows format (i.e. begin with the drive letter).

[skip ci]

 discretionary_policy: fix noprint option

See merge request Dynare/dynare!2349

Output argument and return statement were missing

 fix loop in kalman_update_engine.m if regime_start hits 0

Closes #1956

See merge request Dynare/dynare!2348

Closes #1956

 model_diagnostics.m: set missing variable for bytecode

See merge request Dynare/dynare!2347

[skip ci]

[skip ci]

In Octave, calling @dseries/plot multiple times with `hold on` does
not allow for plotting multiple time series of different lengths on
the same figure (it throws an error).

[skip ci]

[skip ci]