Records used for physical boundary ids are confusing

[Stephan Kramer]

Currently in the assignment of physical boundary ids the physical id is taken from the very first field of the record associated with each shape. For vector layers where shapes have multiple fields this is confusing and error-prone. For instance a different ordering of the fields but with exactly the same data will give different results. In some cases shapefiles will already have some fields associated with them, for instance starting with a .kml shapefile it will already have a 'name' field that is string-valued. Even if you then in qgis add your own 'id' field to it, it will still pick up the 'name' field instead which will lead to a random crash as the plugin doesn't expect string values (this should be handled by a proper error message btw). Also when merging different shapefiles into one the fields will be merged by 'name', so having things rely on order doesn't really make sense.

Suggested fix: have the plugin look up the correct field by looking for the field with the 'id' name.