Create RpcResult class and generalise core special cases

This results in a cleaner MTProtoSender, which now can always
read a TLObject with a guaranteed item, if the message is OK.

telethon/tl/core/__init__.py

@@ -0,0 +1,26 @@
+"""
+This module holds core "special" types, which are more convenient ways
+to do stuff in a `telethon.network.mtprotosender.MTProtoSender` instance.
+
+Only special cases are gzip-packed data, the response message (not a
+client message), the message container which references these messages
+and would otherwise conflict with the rest, and finally the RpcResult:
+
+    rpc_result#f35c6d01 req_msg_id:long result:bytes = RpcResult;
+
+Three things to note with this definition:
+1. The constructor ID is actually ``42d36c2c``.
+2. Those bytes are not read like the rest of bytes (length + payload).
+   They are actually the raw bytes of another object, which can't be
+   read directly because it depends on per-request information (since
+   some can return ``Vector<int>`` and ``Vector<long>``).
+3. Those bytes may be gzipped data, which needs to be treated early.
+"""
+from .tlmessage import TLMessage
+from .gzippacked import GzipPacked
+from .messagecontainer import MessageContainer
+from .rpcresult import RpcResult
+
+core_objects = {x.CONSTRUCTOR_ID: x for x in (
+    GzipPacked, MessageContainer, RpcResult
+)}