Incoming bill actions now have either <committee> or <committees> nodes

Starting on Dec. 13 I started getting 689 bills (in the 116th Congress) with "committees" instead of "committee" in actions in the incoming
GPO bulk data XML files. (Example: [HR 3](https://www.govinfo.gov/bulkdata/BILLSTATUS/116/hr/BILLSTATUS-116hr3.xml)) It looks like currently
1,897 bills have this change. It appears that the data format change described at usgpo/bill-status#147 was published
early, and only in some bills.

Thankfully, our JSON data format for bills already held a list of committee codes, so no change to our data format was needed.

Once GPO has refreshed all of the bulk data files with the new <committees/> format, this code can be simplified to remove the <committee> case.

Fixes #245.

tasks/bill_info.py

@@ -477,12 +477,24 @@ def action_for(item):
 
             references.append({'type': type, 'reference': reference})
 
+    # extract committee IDs
+    if item.get('committee'):
+      # Data format through Dec. 13, 2019 had only one <committee/> (though node could be empty).
+      committee_nodes = [item['committee']]
+    elif item.get('committees'):
+      # Starting on Dec. 13, 2019, and with a slow rollout, multiple committees could be specified.
+      # Thankfully our JSON output format allowed it already.
+      committee_nodes = item['committees'].get("item", [])
+    else:
+      # <committee/> or <committees/>, whichever was present, was empty
+      committee_nodes = []
+
     # form dict
 
     action_dict = {
         'acted_at': acted_at,
         'action_code': item.get('actionCode', ''),
-        'committees': [item['committee']['systemCode'][0:-2].upper()] if item['committee'] else None,
+        'committees': [committee_item['systemCode'][0:-2].upper() for committee_item in committee_nodes] if committee_nodes else None, # if empty, store None
         'references': references,
         'type': 'action', # replaced by parse_bill_action if a regex matches 
         'text': text,