Open PR on deal.II with typos corrections

[luca-heltai]

• In the documentation of Step 26: in the definition of f(x,t) (paragraph Testcase, there's a missing $\tau$

• In the documentation of LinearSolveFunction, SundialsOperator and SundialsPreconditioner are mentioned as arguments, then those are referred to be objects of class LinearOperator. Could this be a typo?

• In the documentation of ARKode, the Detailed Description says that the solve_mass function can be provided to use a custom solver instead of the default one (SPGMR), but the documentation of solve_mass says that this function is mandatory if the mass matrix is not the identity. In other words, SPGMR is not used automatically if solve_mass is not provided, and indeed not providing this function results in a runtime exception related to arkStep_FullRHS.

• In the documentation of ARKode and in the file arkode.h, the function mass_preconditioner_solve has 5 arguments, but the documentation says that there's also a sixth parameter gamma.

[ste-dtst]

Addition: there's a small typo in the documentation of FunctionParser::initialize. It is said that for time-dependent problems one should use FunctionTime::get_time(), but the correct method is FunctionTime::set_time().