L. Lhier's second answer ($g(x) = 1$ and $f(x) = 0$) is probably the minimal one. But you could look at $p$-integrals too. Let $f(x) = \frac{1}{x^2}$ and $g(x) = \frac{1}{x}$. Then $0 \leq f(x) \leq g(x)$ for all $x \geq 1$, and $\int_1^\infty g(x)\,dx$ diverges, but $\int_1^\infty f(x)\,dx$ converges.
Intuitively speaking, an improper integral (at least for a nonnegative function) is attempting to measure the area of an infinite region: the region under the graph, above the $x$-axis, and to the right of $x=a$. If $0 \leq f(x) \leq g(x)$, the region below the graph of $f$ is contained in the region below the graph of $g$. Therefore, if the region below the graph of $g$ has finite area (the integral converges), the region below the graph of $f$ must also have finite area. Contrapositively, if the region below the graph of $f$ has infinite area (the integral diverges), the region below the graph of $g$ must have infinite area. But if the region below the graph of $g$ has infinite area, the region below the graph of $f$ could have finite or infinite area.