* Important bugfix in pulldbf module. Was forcing types on numpy, biggest float
was 'f4'. In case of excel kind of dates for example, this is not enough, and
numbers were lost. The result was repeating equal numbers that should not be
equal.
* Since forcing types on numpy was deprecated, an issue with missing values in a
dbf file (nulls) was easy to fix. Now, null values are replaced with numpy
nan, was (0). Also very important.