Power management can be tricky, especially with chips that have complex powerup schemes like FPGAs. They sometimes require the power rails to be sequenced in a certain order, and so on. I'll assume that's not the case in your application.
Simplest I can think of is to keep the MIC5356 you've found for the 3.3V scenario, but add two low Rds_on logic-level P-channel mosfets (like DMP2035U-7) with source tied to VCC_IN, drain of Q1 tied to IO1 & IO2, and drain of Q2 tied to VCCINT. Tie the gates together, pull them up to VCC_IN with a 47k resistor, and drive them with a voltage monitor like BU4220G or NCP301LSN20T1G (2.0V threshold, active low. 1.9V threshold sometimes also available). The voltage monitor will pull the gates low when the input voltage is below 2.0V - (hysteresis and tolerance), bypassing the LDO and connecting IO1, IO2, and VCCINT to VCC_IN.
That's just an idea. No guarantees you won't break a fancy CPLD chip.
I had to make a supply monitor / cutoff circuit for a work project once. I ended up using a window comparator with internal reference (TPS3700, but MIC841 is cheaper) followed by a PIC10LF320. This allowed me to ignore short glitch periods, to check for undervoltage and overvoltage, and to insert custom time delays. The micro controlled the input load switch (P-mosfet) and both the micro and comparator were supplied from a 3.3V regulator with up to 40V input voltage. The load could operate from 10.8V - 16V, but had to be turned off outside that range.
So I'm biased, but a window comparator with internal reference whose output is checked by a tiny microcontroller is quite useful in these situations. The MIC841 runs on 1.5V - 5.5V Vcc, and the ATTiny9 runs on 1.8V - 5.5V Vcc. Pairing these with a MIC5356 and two bypass P-mosfets would let you delay Pmos turn-on until the supply was within your 1.8V range for a solid second, for example.