Nyquist says that it will have to sample at more than 100kHz

Signal is about 50kS/s

What is kS/s? Is that samples/second? Again I/we know too little.

 

What is kS/s? Is that samples/second? Again I/we know too little.

as written: kilo Siemens per second! the rate of change of conductance? Assumed to be kHz.
first rule of getting the right answer to the question - get your units right!

 

What is kS/s? Is that samples/second? Again I/we know too little.

I would imagine that refers to the ADC sample rate. 50kS/s would be 50 kilo samples per second, or 50000 samples each second.

 

I think the two resistors and one capacitor circuit would work with any supply voltage as long as the supply voltage is as stable as needed for the measurements. If it varies by 5 percent for example then that might be too much.
If the supply voltage is not 3.3v then just use two different values resistors to develop the offset voltage.

Something else to think about is the time delay before the first measurement can be made after power up. The time constant of the circuit would be RT*C where RT is the parallel combination of the two resistors, and because C might be larger then the time delay might be longer. Five time constants is usually enough for settling to within about 1 percent of the final value, but in some ADC applications it may have to be seven time constants for a settling time to within about 0.1 percent of the final value.