230V A.C. mains detection issue, GPIO not stable and false detection due to noise

Hi,
My application is to detect the AC voltage presence using Continuous high or low signal.
I am Getting Same issue with TCA6424A, MSP430F167, Atmega328P and ESP8266
I am using a Non-Isolated power supply with 120mV Vp-p as supply.
Circuit is as shown below where I need GPIO continuous Hi or Low

With this when voltage detected it is not going to zero instead it is somewhere around 0.6-1.2v

And sometimes there is a noise on IO which is Reaching Vp-p as much as 32V as shown below

I Tried Below circuits , with TVS

The 34v noise clamped and I am only getting Ripple of 0.4-0.6v at AC Presence
When I used Schottky arrangement of GPIO As shown below 34v noise clamped but ripple was 0.8-1.2v

now my question is which arrangement is better for my GPIO stability?
TVS of Schottky configuration? Because rise time of noise is less than 10nS. even TVS is giving better performance I am not sure will it sustain those type of continuous noise!
And controller is stucking / not responding when I get this noise or sometimes giving random AC detection signals. but with TVS/ schottky arrangement its not happening but when The GPIO state is above 1v instead of ground Its false triggering like detecting high even there is AC present. is there any way I can make this low level to 0V or less than 200mV

Can I know what type of noise it is? and its source

If there really is a problem of noise, then I think you have two different problems. For the moment let's assume that the noise is a problem with the circuit wiring and/or 'scope grounding. Certainly the high frequency noise that you are seeing is probably not coming through the opto, unless it is faulty. Since you seem to be able to quickly change between micros, it would suggest it is easy enough to remove the micro from the circuit entirely and I suggest you do that. You also need to be looking at the down at the 'scope timebase setting of around 20mS.
You give no value for the resistor connected to the LED of the photocoupler so we don't know what the current is through the LED. But perhaps as important, there is no indication of the power rating of the resistor. If we do a thumbnail calculation, at 10mA RMS the resistor will be dissipating 2.4W. Depending on the resistor it should be getting warm if you used a 5W resistor and positively hot if you used a lower wattage. DO NOT USE YOUR FINGER TO TEST THE TEMPERATURE!!! 240V IS LETHAL. For this kind of circuit I have used a capacitor instead of the resistor. The impedance is (1/2*pi*f*c). For 10mA you would need about 0.15uF, but of a high voltage say 400VDC.

With no micro connected, and the input switch open, what is the output voltage. It should be a steady 3V3. No? look for why not.

Close the input switch. Monitor the output voltage. Is it a steady logic zero. your 4.7K/10u should be fine as a filter. If the voltage changes( in association with the 50Hz signal) to outside the logic low, either there is not enough current going through the LED or your output RC filter is too low. If this is the case, remove the 10uF cap- and investigate the waveform. It should give you an idea if the problem is on the input side.
Once you are sure that the input isolation circuit is working fine- check if the high frequency noise is still there. If yes, either the opto is faulty/wired incorrectly or the noise is coming from the power supply.Once resolved, now you can connect the micro- is the noise present? You could try to suppress the noise with a 0.1u ceramic cap in parallel with the 4.7u.