1), 2) and 3)
If you use different power supplies in a circuit you have to connect them some way or another so that they have a common reference. You will almost always connect grounds, since they are your reference. Voltage is relative: if you take the batteries' plus as a reference the minus will be at -12 V, if you take the minus as reference the plus will be at +12 V. Few circuits will use the plus as reference, we like positive voltages better. So the batteries' minus goes to the Arduino's ground.
Why do they have to be connected? Your transistor will see two currents: a base current, entering the base and returning to the 5V supply through the emitter, and a collector current entering the collector and also returning to the battery via the emitter. Since the currents have the emitter in common (it's called a common emitter circuit) that will be where both power supplies will be connected.
How does the base current know which way to go when it exits the transistor via the emitter? Current can only flow in a closed loop, from the plus from the power supply to the minus. The base current started at the +5 V, so it would not close the loop when it would go the way of the batteries' ground.
4)
We'll leave R2 out for a moment. Because the base-emitter acts as a diode the base is at around 0.7 V. You apply 5 V to activate the transistor, then according to Ohm's Law the current through R1 (which is the base current) is \$\frac{5V-0.7V}{R1}\$. The transistor will amplify that current to a sufficiently high collector current to drive the relay. What's sufficiently high? Therefore you have to check the relay's datasheet. It will either tell you the required current, or the coil's resistance, and then you can calculate the current, again with Ohm's Law. A relay typically needs around 400 mW to activate, so for a 12 V relay that would be a current of 400 mW/12 V = 35 mA. That's the minimum collector current.
To find out how much base current we need to get that we have to look into the transistor's datasheet. Let's say I have 100 000 BC547Bs lying around (I forgot the decimal point when I ordered them) for which I need a purpose. The current gain is given by the \$h_{FE}\$ parameter, which we find on page 2 of the datasheet. For the BC547B that's minimum 200. (Always use worst case values, for \$h_{FE}\$ that's the minimum value. If you use typical values you may have too little current for some parts.)
So to get 35 mA collector current we need 35 mA/200 = 0.175 mA base current. Then R1 has to be \$\frac{4.3 V}{0.175 mA}\$ = 24600 Ω. That's a value you won't find, so shoudl we choose a higher or lower value. If we would pick a higher value the current will be lower, also the collector current will be less, and our relay may not activate. So it has to be lower, the 24600 Ω is the upper limit. Now there's nothing wrong with supplying too much base current (within reason); the collector current will try to follow, but the coil's resistance will limit it. If the coil's resistance is 360 Ω then Ohm's Law says you can't get more than 35 mA at 12 V, no matter how hard you try.
Let's pick a 10 kΩ resistor. That's a much lower value than we needed but we'll be OK. The base current will be around 0.5 mA, which the Arduino will supply happily, and the transistor will try to make that 100 mA, but again, it will be limited to our 35 mA. In general it's a good idea to have some margin, in case the 5 V would be a bit less, or whatever variations there may be else in the parameters. We have a factor three safety margin, which should be OK.
What about R2? We didn't use that and everything seems to be OK. That's right, and it will be in most cases. When would we need it? If the output low voltage of the Arduino wouldn't go below 0.7 V so that the transistor also would get current when off. That won't be the case, but let's say the output low voltage would stick at 1 V. R1 and R2 form a resistor divider, and if we choose R1 = R2 then the 1 V input would become 0.5 V base voltage, and the transistor wouldn't get any current.
We had 0.5 mA base current when on, but with R2 parallel to the base-emitter we'll lose some of that current there. If R2 is 10 kΩ it will draw 0.7 V/10 kΩ = 70 µA. So our 500 µA base current becomes 430 µA. We had a lot of margin, so that would still give us enough current to activate the relay.
Another use for R2 would be to drain leakage current. Suppose the transistor is driven by a current source, like an optocoupler's phototransistor. If the optocoupler sources current it will all go into the base. If the optocoupler is off the phototransistor will still create a small leakage current, what's called "dark-current". Often not more than 1 µA, but if we don't do anything about it it will flow into the base and create a 200 µA collector current. While it should be zero. So we introduce R2, and choose a 68 kΩ for it. Then R2 will create a voltage drop of 68 mV/µA. As long as the voltage drop is less than 0.7 V all current will go through R2, and none into the base. That's at 10 µA. If the current is higher R2's current will be clipped at that 10 µA, and the remainder goes through the base. So we can use R2 to create a threshold. The dark-current won't activate the transistor, because too low.
Except for this case of current-driven R2 will very rarely be necessary. You won't need it here.