Implement *_get_5D(...) array accessors

[atrosinenko]

This PR should technically fix the possibility to at least build models like Buildings.HeatTransfer.Windows.BaseClasses.Examples.AbsorbedRadiation and looks similar to what already exists.

I have rephrased the getIndex_nD function implementation, so they should be easier to read and, possibly, faster to compute (but this highly depends on what the optimizer already does). This may interfere with some corner cases like integer overflow, but these are probably UB-like cases anyway. In case the old implementation was written in such a way for a purpose, then maybe it would be worth to write some comment like

// Index = i0 * dim[1] * dim[2] * ... * dim[n] + i1 * dim[2] * ... * dim[n] + ... + i{n-1} * dim[n] + i{n}

[atrosinenko]

Hmm... There are references to string_get_2D in KeyWordIO.Examples.ReadStringCSV as well...

[atrosinenko]

Adapted real_get_nD for strings, hope this copy-paste is not agains CLA. :)

[hkiel]

Your patch looks good. I wonder if 6D or 7D is also needed in some other model...
I started hudson job to run the testsuite on this and merge if successful.

[OpenModelica-Hudson]

The tests run correctly according to OpenModelica_TEST_PULL_REQUEST 2019-01-03_17-52-50.

[atrosinenko]

Thanks! Meanwhile, on other low-hanging fruits: there are some timeouts like this where according to Callgrind, 60% of instructions are under fmi2GetReal -> oms3::fmi2logger. This can probably be optimized, but I don't know how right now.