[dims] previously non-consecutive dims could get lost

[JanMarvin]

if rows and columns were entirely omitted. This could create a cc object that was to short. fixes #1014

[JanMarvin]

It's not the proudest thing I've ever programmed, but it should be better than the previous approach.

The problem is that the length of the dims of the data to be selected and the entire data range are unequal.
Imagine a case where dims skips a row, as in “A1:B2,A4:B6”. In some places, we simply loop over the number of columns and rows. If a row is missing, our loops are not aware of it. There is nothing to do for row 3, but they iterate over it and are at n+1 and therefore later stop at n-1. The correction therefore enlarges the selected data frame so that it covers the entire data range. This way, when looping over the data, both sizes are equal.

Unfortunately, in these cases we might have to create a few empty columns and rows.

The whole code is unfortunately a little ... confusing. I'm not going to call it spaghetti code just yet, but it's getting close due to the various stages of development we're still supporting. One day write_data() and write_data2() should be combined and the whole thing tidied up. We can also iterate over the necessary rows and columns. But that will probably only happen in the distant future.