more robust parsing of user counts

The latest userstats-relay-country.csv file has a row with scientific
notation. Parsing as a float first prevents failing on that.

Author: jtracey

tornettools/stage.py

@@ -60,7 +60,7 @@ def stage_users(args, min_unix_time, max_unix_time):
 
             date = str(parts[0]) # like '2019-01-01'
             country_code = str(parts[1]) # like 'us'
-            user_count = int(parts[2]) # like '14714'
+            user_count = int(float(parts[2])) # like '14714' or '2e+05'
 
             dt = datetime.strptime(date, "%Y-%m-%d").replace(tzinfo=timezone.utc)
             unix_time = int(dt.strftime("%s")) # returns stamp like 1548910800