I am reading the "Quantum Field Theory lectures of Sidney Coleman". In the first chapter (subchapter 1.2), the author talks about translation invariance. In particular, he states that $U(a)=e^{iP\cdot a}$ is the translation operator, with $$O(x+a)=U(a)O(x)U^{\dagger}(a)$$ where $O$ is any other operator depending on $x^{\mu}$. Then, the author reduces the translations to space translations and therefore $$e^{-i\textbf{P}\cdot\textbf{a}}O(\textbf{x})e^{i\textbf{P}\cdot\textbf{a}}= O(\textbf{x}+\textbf{a}).$$ But then, the author states that only operators localized in space transform according to this rule, giving a counter-example the operator $\hat{\textbf{q}}$ $$e^{i\textbf{P}\cdot\textbf{a}}\hat{\textbf{q}}e^{-i\textbf{P}\cdot\textbf{a}}= \hat{\textbf{q}}+\textbf{a}.$$
I have three questions:
What is the underlying reason for only operators localized in space being transformed under this rule?
Why is $\hat{\textbf{q}}$ an operator that is not localized in space?
How do the operators that are not localized in space transform?
Any help will be appreciated.