Details about difficulty adjustment can be found at https://en.bitcoin.it/wiki/Difficulty and https://en.bitcoin.it/wiki/Target. Difficulty adjusts every 2016 blocks, with the goal of keeping the average block time at 10 minutes (so that 2016 blocks takes 2 weeks). At each adjustment, the difficulty can change by at most a factor of 4 (up or down).
Let's assume the worst case: that this disruption happens immediately after a difficulty adjustment. With a 90% drop in mining power, each block now takes an average of 100 minutes to mine, instead of the usual 10. So the next 2016 blocks take 20 weeks instead of 2.
After 20 weeks, there is an adjustment, and presumably difficulty will drop by the maximum factor of 4. Now the average block time is 25 minutes and the next 2016 blocks will thus take 5 weeks.
At this point the difficulty can drop by a factor of 2.5 and we are back on track, with an average block time of 10 minutes.
So it takes 25 weeks to get back to normal.
Of course, in your situation, a more likely outcome is that Chinese miners will quickly sell their mining hardware to people outside the country, where electricity is still plentiful, and the total mining power will recover. (Or they'll set up their own offshore mining facilities, and just ship the hardware there.) If not, then mining hardware producers would try to quickly ramp up production so as to fill the void. After all, everyone can see that difficulty will drop by a factor of 4 in 20 weeks, which will increase mining revenue by a factor of 4 and make it far more profitable (assuming the price of Bitcoin remains constant).