Do I just connect thermal pads to the ground plane and call it a day?

Yup.

Wouldn’t that make the components hard to solder with hot air?

Sorry, I’ve never tried hot air assembly.

Do I make an isolated polygon that only acts as a thermal pad?

Ideally the copper area is big to spread out the heat. If you have an isolated polygon it can’t spread very far and buys you less cooling.

The 7333A is a linear regulator, which means it drops voltage by converting power to heat. Typically those make sense when the input voltage is close to the output voltage or the load is very small. If it’s getting too hot, the load is high enough that the efficiency will be very bad…whether or not this is a problem depends on your application.

Some random site claims 170mA and another claims up to 400mA. 170mA * 8.7V (12V in minus 3.3V out) = about 1.5 watts, which is too much for a TO-92 package.

Can you use a tiny buck converter instead? Or a larger package for the linear regulator that can add a small heat sink?

As for your actual circuit, the second transistor is an interesting idea (you’re using it to invert the state so you can have the GPIO pulled in the non-problematic direction?) and I don’t have enough experience to give further suggestions.

I’m not entirely clear on the problem, but yes - the circuit as drawn makes the microcontroller pin start high, then fall after some time. Do you need the microcontroller pin to have a different voltage than the transistor base (I assume when you said gate you mean base…gates are for FETs), or is this good enough?

Would a circuit like this power-on reset circuit work for your application?

LFP cells have excellent cycle life anyway (2000+ cycles); is it worth worrying about staying at 95%?