IMHO the question, and hence answers, have a significant potential to be misleading.
It is necessary to consider how to power an IC (via Vcc and GND pins). However, it is not sufficient to ensure an entire circuit will work.
As well as power and ground, the vast majority of (Vcc/GND powered) ICs also have input and output signals. These signals must be considered too as part of a practical solution.
Typically, a signal will need either a shared voltage reference, and/or a common current path. Putting resistors into ground and power could significantly compromise the signals integrity.
For example, consider an output signal from the IC. The 'load' being driven may be sensitive to the voltage driving it, or the current flow. If its ground is tied directly to the negative side of the power supply it may input the (dangerously) wrong input voltage. If its 'ground' must be tied to the same ground as the IC, then its downstream inputs may also need to be tied to the same ground. However, this might be impossible, because it is separate equipment.
Further, that load may need current, disturbing the GND reference voltage. For multiple IC outputs this may be unacceptable.
The inputs to the IC will likely have requirements to, which may also be difficult to satisfy without common ground and voltage references.
So, IMHO, options two and three may be adequate to power the IC, however they are poor places to start when considering the whole circuit's signal requirements.
It may even be that option one is not appropriate, but the analysis and solution design will likely be easier starting from such a simple starting point.
Put another way; if I asked someone to design a way to power a circuit, and option two or three were presented when one was adequate, I would ask for a good analysis for all of the inputs and outputs to demonstrate the alternative was at least as good, and understood by us all. Similarly if they presented option two or three, even if it were adequate, without a good analysis covering inputs and outputs, I would ask for that analysis.
IMHO option two and three imply sufficient extra complexity that we'd need to have a documented analysis to identify and explain the assumptions and benefits of those approaches. If there are no clear benefits, we'd revert to option one. The common understanding of option one would help us develop, test, interface and evolve the circuit. If we departed from option one, we'd be clear why.
IMHO, it is unlikely that we would use option two or three to overcome practical problems caused by using option one.