If you are trying to do it the "correct" way, you'll get cumbersome schemes, I've even seen some in which both sides of the heating element are isolated with two MOSFETs, so the thermocouple's voltage can be properly read. This is even more hilarious when we think there is no cold junction compensation for the other side of the thermocouple, which makes the whole thing about 10-30 Celsius inaccurate.
However, if you search for schematics in existing products, most of them are using a LM358 and a series resistor, sometimes with diodes tied to the V+ and V- of the op-amp power supply, but not necessarily. The capacitor from the input of the op-amp to GND seems to be an essential component, too.

How could that measure tens of microvolts, with 24V AC or PWM over it, with enough precision and accuracy over 1-2 meters long wires

, it completly puzzles me.

No need for expensive op-amps or low noise and auto offset compensated chopping op-amps.
The LM358 seems to be the magic part that just works.
