I think it best to write the functional derivative unambiguously as the explicit expression
$$
\delta S = -\frac 12 \int_M \delta g^{\mu\nu} T_{\mu\nu}\sqrt{g}d^dx,
$$
where $S$ is the (scalar) action functional. Then, as $\delta g^{\mu\nu}$ is a tensor, and $\sqrt{g}d^dx$ an invariant, it is clear that $T_{\mu\nu}$ is a tensor.
The potential ambiguity is whether to include the $\sqrt{g}$ in the measure in the integral definition of $\delta S/\delta g^{\mu\nu}$ or not.
I just noticed that I had answered this a couple years ago:
Is $T^{\mu\nu}=-\frac{2}{\sqrt{-g}}\frac{\delta S_m}{\delta g_{\mu\nu}}$ a true tensor or a density?