The Paley-Wiener criterion defines a condition on the magnitude spectrum of a causal time-domain function. So if the Paley-Wiener criterion is satisfied for a given $A(\omega)=|H(\omega)|$, we know that there is a causal function with magnitude spectrum $A(\omega)$. It should be noted that the Paley-Wiener criterion is only applicable to square-integrable functions $A(\omega)$. If $A(\omega)$ is not square-integrable then the criterion is neither necessary nor sufficient.
A frequently occurring misunderstanding is that, given a frequency response $H(\omega)$, we can check the causality of the corresponding system using the Paley-Wiener criterion. This is generally not the case. If the criterion is not satisfied we know for sure that $H(\omega)$ is not the frequency response of a causal system. But if it is satisfied we only know that there must be a causal function with magnitude spectrum $A(\omega)=|H(\omega)|$, but we still don't know if the given $H(\omega)$ corresponds to a causal system. This depends on the phase of $H(\omega)$, and the only thing we know is that it is possible to find a phase response $\phi(\omega)$ such that $H(\omega)=A(\omega)e^{j\phi(\omega)}$ corresponds to a causal system.
To answer your questions:
- Yes, it only applies to LTI systems because only LTI systems are fully characterized by a frequency response.
- Yes, only causal systems can be realized because we cannot look into the future.
- This question is a bit broad. There are realizability conditions for certain restricted sets of systems (e.g., passive LTI systems), but in general, causality is definitely a necessary condition. However, for the above mentioned reason we cannot just take a given $H(\omega)$, check the Paley-Wiener criterion and conclude that we can realize $H(\omega)$ as a causal system, because we didn't take into account the system's phase.