Fix subscribe qos fixed header value

[picaoao]

MQTT protocol states that qos must be set to 1 in the fixed header of the subscribe frame. Although it was not being a problem when using with emqttd, vernemq was not accepting the frame.

[tisba]

@picaoao do you have a reference for that?

I'm not an mqtt expert, but reading http://docs.oasis-open.org/mqtt/mqtt/v3.1.1/os/mqtt-v3.1.1-os.html#_Toc384800437 it looks like the fixed header is indeed fixed and does not contain the QoS flag at all.

Don't get me wrong, I'm just curious to understand the protocol a bit better :)

[tisba]

@picaoao FYI I just changed the target branch to develop.

@nniclausse This seems to be a simple change, but I'm not very familiar with MQTT. Are you? Or do you know how we can figure out a reference that this changes makes sense?

[nniclausse]

@tisba: looking at the mqtt_frame code, it seems that the fixed_header built by tsung is indeed wrong. So this patch may fix the problem, but by side effect.
encode_fixed_header should handle subscribe packets differently. The second byte is fixed and always 0010, dup, retain and qos are meaningless in this case.

[tisba]

#327 also addresses the changes in this PR and in addition fixes the subscribe control packet as @nniclausse also pointed out.

I'm closing this in favour of #327.