Dynare will by default generate the IRFs to a one standard deviation shock. In your case, the standard deviation is 0.1. If m is measured in percent, then the shock will be 10%. The interpretation of the IRF of c depends on the units in which c is measured. Regarding size: your are the model builder and are the only one who can have an idea what to expect from the model.

I have a similar question related to the shocks in the oo_.exo_simul structure. How do we have to interpret those shocks when we have a log-linearized model, i.e are the shocks also log-linearized or are they the absolute, non-linear shocks?

In addition to that:
In what measure are they expressed? Are these shocks given as a stderr, absolute numbers or percentages?

I am not sure I understand the question. The interpretation of shock will depend on the way they enter the model. Additive shocks to logged variables have the interpretation of percentage shocks. Additive shocks to levels are in the same units as the variable you add them to.

Yes, that answers my question.
And the same applies to variables from the oo_.endo_simul structure, right?
For example, in my log-linear model I have a value for mg in the oo_.endo_simul structure of 0.0345 in t=25. That means mg responds to all shocks in t=25 with an increase of 3.45%, correct?