With circuit switching, you have a 1-to-1 circuit with some predefined channel capacity which is guaranteed. For it to work, you must first establish a circuit/connection before you send any data, and break the connection after.
With packet switching, it's somewhat like the local post office - you bring the packets, send them, and let the other devices forward them around, to one or many different destinations.
What you mentioned in the first part is called multiplexing. When you establish, for example, a phone call, you get a 1-on-1 connection with the other side, and full channel capacity — 0–3400kHz for voice (at least in some parts of Europe), which only you can use. Since using separate wires is expensive, different multiplexing technologies have appeared, such as TDM, where your channel is "compressed", but you still have a 1-to-1 connection with the same capacity. With FDM, your call is mixed up to higher frequencies, but you still get a 1-to-1 channel, with the same capacity.
Yes, the medium is shared, but you still get the full capacity you have payed for, and a 1-to-1 channel.
As to the second part, circuit switching requires a circuit to be set up (you have to dial a number), and after sending the data/voice, you have to break down (shut down) the connection. When you need to communicate to a a lot of different people/devices, this takes time, since when one connection (call) is active, you cannot use that line for anything else, even if you are still waiting for the other person to answer/recieve the data. If compared to traditional networks, it takes a lot less time to send a 100 letters by post, then calling 100 different people.
Another problem with circuit switched systems is that the channel is used up even when no data is sent through it (noone else can use that channel at that time). If you have an 8-line telephone uplink, and even if all 8 callers are waiting on hold, your channel is still used up. With packet switched networks you can even oversell your connection (due to statistical multiplex) — if you're an ISP and have a 8Mbit/s uplink, you can sell 10, 20, maybe even more 1Mbit/s connections, since it is statistically improbable, that all of the users will use the connection fully at the same time. And even if they do, their connection still works, with lower speed (compared to blocked/dropped calls on circuit switched networks).
A message could be circuit switched or packet switched. You would probably classify message switching as an application layer construct.
Packet switching can break a large message into smaller packets to be sent, but a message may be small enough to fit into a single packet, too.
A circuit could carry a message over a circuit switched network.
It's kind of like comparing peer-to-peer routing with IP routing. They are not equivalent since one is application layer routing, and the other is layer-3 routing on which the application layer routing may happen.
Best Answer
This seems to assume that queues can have limited number of packets (true enough), but unlimited packet size (not true). The idea that a large message packet size uses fewer places in the queue is true, but it ignores reality that an interface can only serialize the packet bits at a certain speed (bandwidth), so that larger packets take proportionally longer to transmit, stalling the queue for a longer period of time.
The bottleneck is the bandwidth of the interface. If an interface can transmit at a limited speed (true), the number of bits in the queue will be the same, regardless of the packet size. What is true is that the larger message packets will be fewer in the queue, but will still take about the same amount of time as smaller packets to transmit the same amount of data because the interface can only serialize the bits at its fixed speed.
Transmitting 1 Gb through a 100 Mbps interface takes the same amount of time regardless of the packet size. The only thing larger packets gain is eliminating some packet header overhead, but that can be ignored unless the packets are so small that something like the IPv4 20-byte packet header is a substantial percentage of the packet size.
In any case, the real world uses packet switching. Circuit switching, such as the traditional telephone circuit-switching network, is going away in favor of packet switching because packet switching is more flexible, allowing the circuit to be shared more easily and allowing different services, e.g. voice, data, video, etc., to use the same circuit at (relatively) the same time.