HTTP_Upgrade: new exception

The previous content was confusing, outdated, and unhelpful.

Thanks, the fix looks good. I forgot to port notify_on_close to linux during the process.  

read_thread(): catch exit sentinel
request.close(): push exit sentinel
response_re, _read_message(): collect 'reason' (though it's useless)
_read_message(): use req.has_body()
send_request(): pass in extra args to request() constructor

HTTP_Protocol_Error: forgot to move this from server.py
http_file.__init__(), http_file._gen_read_all() : handle the old-school situation where data goes until "connection: close"

added: enum for FLUSH values

header_set.__delitem__(): remove a header
header_set.copy(): copy a header set

Bad_Response: new exception

server.start: set the thread name based on self.__class__
server.create_connection: allow the user to override the connection class (for spdy)
tlslite_server.__init__, accept: allow the user to pass handshake_args down (for spdy)

worm.kill(): remove from arena.worms at this point, rather than relying on the try/finally in the thread.

Conflicts:
	coro/_coro.pyx

buffer.set16(): method to set a 16-bit length field after the data is emitted
Packer.addbyte(): call underlying buffer add_byte().
Packer.addstring(): chr() call not needed
Packer.endRR(): call self.buf.set16() to set RR length
Unpacker.getstring(): ord() call not needed
Unpacker.getHINFOdata(): useless, but since addHINFOdata() is there...
Unpacker.unpack(): cut & paste typo