Let's first look at a fixed $n$. For $x \in \bigl[\frac{k}{n}, \frac{k+1}{n}\bigr)$ we have
$$T_n f(x) = \int_0^{k/n} f(t)\,dt,$$
and therefore
$$Tf(x) - T_nf(x) = \int_{k/n}^x f(t)\,dt.\tag{1}$$
To get a nice bound for $\lvert Tf(x) - T_nf(x)\rvert$ we use the Cauchy-Schwarz inequality,
$$\lvert Tf(x) - T_nf(x)\rvert \leqslant \sqrt{\int_{k/n}^x 1^2\,dt} \cdot \sqrt{\int_{k/n}^x \lvert f(t)\rvert^2\,dt} \leqslant \sqrt{x - \tfrac{k}{n}} \cdot \lVert f\rVert_{L^2([0,1])}.\tag{2}$$
Since $x - \frac{k}{n} < \frac{k+1}{n} - \frac{k}{n} = \frac{1}{n}$, we see that $\lvert Tf(x) - T_nf(x)\rvert \leqslant \frac{1}{\sqrt{n}}\lVert f\rVert_{L^2([0,1])}$ for all $x \in [0,1]$ (the equality $T_nf(1) = Tf(1)$ follows directly from the definitions). Thus we find
$$\lVert (T - T_n)f\rVert_{L^2([0,1])}^2 = \int_0^1 \lvert Tf(x) - T_nf)(x)\rvert^2\,dx \leqslant \int_0^1 \frac{1}{n}\lVert f\rVert_{L^2([0,1])}^2\,dx = \frac{1}{n}\lVert f\rVert_{L^2([0,1])}^2,$$
from which we obtain $\lVert T - T_n\rVert \leqslant n^{-1/2}$ by taking the square root on both sides, and then the supremum over all $f$ with $\lVert f\rVert_{L^2([0,1])} \leqslant 1$. (Aside: if we had used the right hand side of $(2)$ rather than the bound $\sqrt{x - \frac{k}{n}} \leqslant \frac{1}{\sqrt{n}}$ we would have obtained the slightly sharper bound $\lVert T - T_n\rVert \leqslant \frac{1}{\sqrt{2n}}$.)
Now $T_n$ is a continuous linear operator with finite-dimensional range - the range of $T_n$ is spanned by the characteristic functions of the intervals $\bigl[\frac{k}{n}, \frac{k+1}{n}\bigr)$, $0 \leqslant k < n$ - hence compact. And by $\lVert T - T_n\rVert \to 0$, $T$ is the norm-limit of a sequence of compact operators. Since the subspace of compact operators is closed, it follows that $T$ is compact too.