[#286,bugfix]: fix for case where fields array has nulls in it.

If passed fields = [ null, ..., ...] or [ ... , null, ...] we fail in Dataset._normalizeRecordsAndFields method because:

1. If null is first element we do not do field generation correctly because typeof(null) is object - ede211646a/src/model.js (L108)
2.  We call toString on field names which fails for null ede211646a/src/model.js (L113)

src/model.js

@@ -105,12 +105,16 @@ my.Dataset = Backbone.Model.extend({
     } 
 
     // fields is an array of strings (i.e. list of field headings/ids)
-    if (fields && fields.length > 0 && typeof(fields[0]) != 'object') {
+    if (fields && fields.length > 0 && (fields[0] === null || typeof(fields[0]) != 'object')) {
       // Rename duplicate fieldIds as each field name needs to be
       // unique.
       var seen = {};
       fields = _.map(fields, function(field, index) {
-        field = field.toString();
+        if (field === null) {
+          field = '';
+        } else {
+          field = field.toString();
+        }
         // cannot use trim as not supported by IE7
         var fieldId = field.replace(/^\s+|\s+$/g, '');
         if (fieldId === '') {