[BUG] Integer QM charge check too stringent

[lohedges]

The integer charge check in the sire.qm sub-package is a little too stringent. This should probably be reduced to something like 1e-2. In addition, we need to be able to redistribute the charge for isolated QM regions, e.g. small molecules. At present, the code only works when the QM region is part of a larger entity.