Optionally represent monetdb str column as fixed with numpy string array

[gijzelaerr]

what we can also think of having an option to format monetdb string columns as numpy fixed-width columns in the result. I know database people find this ugly, but this is just how numpy works and what makes it fast (in the context of python). A common use case in data science panda land is that you have a short string column, containing a dataset label for example. Fixed width strings then suddenly start to make sense again. The issue now is that if you have a big table with int and floats and you call a .fetchdf() monetdbe is fast, but as soon as one of the columns is a (short) string, performance plummets, since it falls back to the python processing mode (with a warning). We could make this a configurable option for .fetchdf() and .fetchnumpy(), where you indicate if you need speed or are memory limited.