BUG: Fix dataset parsing for Hyrax catalogs (Fixes #759)

Due to the way parsing is done, we look backwards (on when a dataset tag
is encountered) to handle datasets with embedded access tags.
Unfortunately, this left us not handling the last such dataset
encountered if it contained such tags, like is done on NASA's Hyrax
server.

src/siphon/catalog.py

@@ -314,12 +314,11 @@ def __init__(self, catalog_url):
                 current_dataset = child.attrib['name']
                 self._process_dataset(child)
 
-                if previous_dataset:
-                    # see if the previously processed dataset has access elements as children
-                    # if so, these datasets need to be processed specially when making
-                    # access_urls
-                    if self.datasets[previous_dataset].access_element_info:
-                        self.ds_with_access_elements_to_process.append(previous_dataset)
+                # see if the previously processed dataset has access elements as children
+                # if so, these datasets need to be processed specially when making
+                # access_urls
+                if previous_dataset and self.datasets[previous_dataset].access_element_info:
+                    self.ds_with_access_elements_to_process.append(previous_dataset)
 
                 previous_dataset = current_dataset
 
@@ -346,6 +345,11 @@ def __init__(self, catalog_url):
                     service_skip = self.services[-1].number_of_subservices
                     service_skip_count = 0
 
+        # Needed if the last dataset had such info, since it's only processed looking backwards
+        # when a new dataset is encountered.
+        if previous_dataset and self.datasets[previous_dataset].access_element_info:
+            self.ds_with_access_elements_to_process.append(previous_dataset)
+
         self._process_datasets()
 
     def __str__(self):