Why is it more typical to place the load in the collector rather than the emitter of the transistor?
The reason is that if there is any resistance between the emitter and ground, the current through the emitter (both base and collector current) cause a voltage drop which increases the voltage which must be applied to the base. In essence it is negative feedback. That is all wonderful in an audio amplifier because it allows differences in gain between various transistors to be evened out (amongst other things). But for a switching circuit we would generally prefer that the transistor act as a switch rather than a linear amplifier, so reducing the gain is something we don't want to do.
Is this only for NPN transistors?
No it applies to PNP as well, and to mosfets -- albeit with a change in the names of the elements.
Why should the PIR be connected with a resistor in series to the base of the transistor? In order to limit current? Is there any reason you picked a 470R?
We don't know what the output of the PIR is. If it is a stiff voltage source, then it may try to maintain 3.3V (or whatever) under load. Given that there is now no resistance in the emitter, all you're looking at driving is the BE junction of the transistor. Without current limiting you may damage either the PIR output or the transistor.
470R just seemed like a reasonable value. You probably could use anything from 100R to 4k7. The actual value you'd choose (if you went to the effort of calculating it) would depend on the nature of the PIR's output, the HFE of the transistor, and the current required for the relay.
470R seemed like a safe value
How would you decouple the PIR and the motor? Use a capacitor?
You would measure the current required for the PIR and place a resistor in series with it that limits the voltage drop to maybe 1V. You would then place a capacitor across the PIR chosen so that the time constant was significantly longer than the duration of any spikes that the motor might place on the power rail. Practically you might just pick 100uF or something like that.
There are other methods. If the PIR required (or could operate from) a lower voltage, then a zener diode could also be used to provide a more regulated voltage.