Ideally, you would like something that is 0-5V (rail to rail). Take a look at application #2. Look at the conversion times and the effects on what you have to read and the accuracies.
It doesn't really matter. I used a 30 A sensor because I didn't know what your after. You said a 40 W bulb. If this is incandescent, you have to account for the resistance being about 1/10 or 1/15 of what it would be for 40 W. Then you have to scale the RMS value x 1.414, so if it were say 1 A, you would be looking at +-1.414A if my math is right.
0 Amps is Vcc/2. Application #2, uses R1 and R2 to generate Vcc/2.
You want to make sure that your A/D can use an external reference. That external reference must be Vcc/2. It should not be a 2.5V reference. It really does need to be Vcc/2 created by either resistors or a real splitter.
A 30A part is way too big for a 40 W bulb. You know that, right? That's like 0.2 Amps RMS. But remember what I said and I'll let you figure it out. You need +- Amps or 0 to the +peak and 0 to the - peak AND you need to account for, or should account for, the temperature coefficient of tungsten. 10 to 15x less in resistance is just a "rule of thumb".
So, what you will get is an approximate 0-Vcc signal where zero is you max full scale -peak current and Vcc is your highest peak current and Vcc/2 is zero current.
If you shoot for something like 0.5 to 4.5 V at your designed currents you have room to play.
Since your reference is Vcc/2, that makes the system "ratiometric". Now it would be really convenient, if you could measure Vcc/2 and the output of the sensor with two successive conversions.