Fix `simult_.m` to call specialized simulation codes for order <= 3 instead of...

See merge request Dynare/dynare!2286

Fix `simult_.m` to call specialized simulation codes for order <= 3 instead of `k_order_simul` whenever possible

Ref. #1859

The test condition was incorrect (though this was probably harmless).

Also the error message related to oo_.dr.ghu was incorrect.

By symlinking ~/.MathWorks to /dev/null, MATLAB is unable to copy and launch
the ServiceHost binary, which seems to cause lockups. See for example:
https://git.dynare.org/Dynare/dynare/-/jobs/62558
https://git.dynare.org/Dynare/dynare/-/jobs/62547

Bug fix in PAC/MCE expectation with cherrypick (was only affecting PAC models with transformed variables).

The RHS of the equation was not acknowledging the user provided name
for the auxiliary variable.

Closes: #1927

Closes: #1721

🐛 make sure user-defined warning states are not overridden

See merge request Dynare/dynare!2285

warning_config.m should only be called once in dynare.m

[skip ci]

[skip ci]

[skip ci]

The initial guess was set to the solution, so the algorithms would really be
tested.

Function Interpreter::check_for_controlled_exo_validity() would give incorrect
results because the sconstrained_extended_path structure was containing 1-based
variable indices (it is expecting 0-based ones).

This prevents a name clash with some function-local variables also named “b”.

manual: clarify treatment of missing data in Inversion Filter

See merge request Dynare/dynare!2284

Replaced

         ᵐ                              ᵐ⁻¹ ᵐ⁻¹
  Zₜ = - ∑  𝛼ᵢ 𝛽ⁱ⁺¹ Zₜ₊ᵢ + A(1) [ 𝛥 yₜ + ∑   ∑   𝛼ⱼ₊₁𝛽ʲ⁺¹𝛥 yₜ₊ₖ ]
        ᵢ₌₁                             ₖ₌₁ ⱼ₌ₖ

by

         ᵐ                              ᵐ⁻¹ ᵐ⁻¹
  Zₜ = - ∑  𝛼ᵢ 𝛽ⁱ⁺¹ Zₜ₊ᵢ + A(1) [ 𝛥 yₜ - ∑   ∑   𝛼ⱼ₊₁𝛽ʲ⁺¹𝛥 yₜ₊ₖ ]
        ᵢ₌₁                             ₖ₌₁ ⱼ₌ₖ

Also:

 - Added comments and comparison with the formula in Brayton et alii (2000),
 - Removed unused input argument,
 - Removed unused variables.