The way I see it the watchdog reset should be used in the main code because if for any reason the code enters a situation with an endless loop (bu some code mistake or some unknown factor) the watchdog will not be cleared and the mcu will reset.
On the other hand since the code in the main stops in order for the interrupt to be executed and then resumes again this would mean that if you clear the watchdog inside the interrupt then the mcu will never get out of this situation , the interrupt will execute and will return to the stuck loop again.