Move all the raw tweets for tests into their own JSON files

Seems better to have the raw data as JSON, like it comes from the API,
then load it into python objects for each test.

tests/tweets/compat.json

@@ -0,0 +1,51 @@
+{
+  "contributors":null,
+  "truncated":true,
+  "text":"Say more about what's happening! Rolling out now: photos, videos, GIFs, polls, and Quote Tweets no longer count tow\u2026 https://t.co/SRmsuks2ru",
+  "is_quote_status":false,
+  "in_reply_to_status_id":null,
+  "id":777915304261193728,
+  "favorite_count":13856,
+  "source":"<a href=\"http://twitter.com\" rel=\"nofollow\">Twitter Web Client</a>",
+  "retweeted":false,
+  "coordinates":null,
+  "entities":{
+    "symbols":[
+
+    ],
+    "user_mentions":[
+
+    ],
+    "hashtags":[
+
+    ],
+    "urls":[
+      {
+        "url":"https://t.co/SRmsuks2ru",
+        "indices":[
+          117,
+          140
+        ],
+        "expanded_url":"https://twitter.com/i/web/status/777915304261193728",
+        "display_url":"twitter.com/i/web/status/7\u2026"
+      }
+    ]
+  },
+  "in_reply_to_screen_name":null,
+  "id_str":"777915304261193728",
+  "retweet_count":14767,
+  "in_reply_to_user_id":null,
+  "favorited":false,
+  "user":{
+    "name":"Twitter",
+    "screen_name":"twitter"
+  },
+  "geo":null,
+  "in_reply_to_user_id_str":null,
+  "possibly_sensitive":false,
+  "possibly_sensitive_appealable":false,
+  "lang":"en",
+  "created_at":"Mon Sep 19 17:00:36 +0000 2016",
+  "in_reply_to_status_id_str":null,
+  "place":null
+}