1. it's intention is provide controlled positive feedback. If you can hiold the circuit just below the point of oscillation it will have better sensitivity and selectivity. The variable resistor is to 'short out' some of the feedback to allow you to set the regeneration level. However, this is not a good way to do it because as the resistor value decreases it also reduces the 'Q' factor of the tuning coil and makes it LESS sensitive. A better method is to control how much feedback is allowed to reach the regeneration coil rather the shorting out what is already there.
Also, it relies on the 100nF capacitor being of poor quality because a good one would short out the signal anyway and no RF would remain to be fed back!
The second and third schematics are much better although both could still be improved.
2. The transistor type is not critical but you should be aware that in a simple bias circuit like that, the amount of current through the transistor may vary considerably, even from one transistor to another of the same type. Do not be tempted to use a lower resistance loudspeaker because it may increase the current beyond the transistors limits.
Brian.