Split stdlib into Python 2 and 3 versions (#5442)

All new files in stdlib/@python2 are straight copies of the
corresponding files in stdlib.

stdlib/@python2/binascii.pyi

@@ -0,0 +1,50 @@
+import sys
+from typing import Text, Union
+
+if sys.version_info >= (3, 0):
+    # But since Python 3.3 ASCII-only unicode strings are accepted by the
+    # a2b_* functions.
+    _Bytes = bytes
+    _Ascii = Union[bytes, str]
+else:
+    # Python 2 accepts unicode ascii pretty much everywhere.
+    _Bytes = Text
+    _Ascii = Text
+
+def a2b_uu(__data: _Ascii) -> bytes: ...
+
+if sys.version_info >= (3, 7):
+    def b2a_uu(__data: _Bytes, *, backtick: bool = ...) -> bytes: ...
+
+else:
+    def b2a_uu(__data: _Bytes) -> bytes: ...
+
+def a2b_base64(__data: _Ascii) -> bytes: ...
+
+if sys.version_info >= (3, 6):
+    def b2a_base64(__data: _Bytes, *, newline: bool = ...) -> bytes: ...
+
+else:
+    def b2a_base64(__data: _Bytes) -> bytes: ...
+
+def a2b_qp(data: _Ascii, header: bool = ...) -> bytes: ...
+def b2a_qp(data: _Bytes, quotetabs: bool = ..., istext: bool = ..., header: bool = ...) -> bytes: ...
+def a2b_hqx(__data: _Ascii) -> bytes: ...
+def rledecode_hqx(__data: _Bytes) -> bytes: ...
+def rlecode_hqx(__data: _Bytes) -> bytes: ...
+def b2a_hqx(__data: _Bytes) -> bytes: ...
+def crc_hqx(__data: _Bytes, __crc: int) -> int: ...
+def crc32(__data: _Bytes, __crc: int = ...) -> int: ...
+def b2a_hex(__data: _Bytes) -> bytes: ...
+
+if sys.version_info >= (3, 8):
+    def hexlify(data: bytes, sep: Union[str, bytes] = ..., bytes_per_sep: int = ...) -> bytes: ...
+
+else:
+    def hexlify(__data: _Bytes) -> bytes: ...
+
+def a2b_hex(__hexstr: _Ascii) -> bytes: ...
+def unhexlify(__hexstr: _Ascii) -> bytes: ...
+
+class Error(ValueError): ...
+class Incomplete(Exception): ...