Version 1.
$$\int_{0}^{1}\frac{\log(x+\sqrt{1-x^2})}{x}\,dx = \int_{0}^{\pi/2}\log(\sin\theta+\cos\theta)\cot(\theta)\,d\theta \tag{1}$$
by enforcing the substitution $\theta\to\frac{\pi}{2}-\theta$ and averaging turns out to be equivalent to
$$ \int_{0}^{\pi/2}\frac{\log(\sin\theta+\cos\theta)}{2\sin\theta\cos\theta}\,d\theta\stackrel{\theta\mapsto 2\arctan u}{=}\int_{0}^{1}\frac{\log(1+2t-t^2)-\log(1+t^2)}{2t(1-t^4)}\,dt\tag{2} $$
which can be managed by partial fraction decomposition, through the dilogarithm functional identities $(3)-(7)$, since
$$ \int\frac{\log(1-t)}{t}\,dt = C-\text{Li}_2(t).\tag{3} $$
The same applies is we avoid the initial symmetrization, since
$$ \int_{0}^{\pi/2}\frac{\log(\sin\theta+\cos\theta)}{\tan\theta}\,d\theta=\int_{0}^{1}\left[\log(1+2t-t^2)-\log(1+t^2)\right]\frac{1-t^2}{t(1+t^2)}\,dt .\tag{4}$$
Version 2. By immediately substituting $\theta=\arctan u$ in $(1)$, the original integral is converted into
$$ \int_{0}^{+\infty}\frac{\log(1+u)-\frac{1}{2}\log(1+u^2)}{u(1+u^2)}\,du $$
which by Feynman's trick equals
$$ \int_{0}^{1}\frac{\pi+2a\log a}{2(1+a^2)}\,da +\frac{1}{4}\int_{0}^{1}\frac{\log a}{1-a}\,da=\frac{\pi^2}{8}-\frac{\pi^2}{48}-\frac{\pi^2}{24}=\color{blue}{\frac{\pi^2}{16}}.\tag{5}$$
(Poly)logarithmic integrals always are a tricky thing, one never knows in advance what is the best moment for enforcing a substitution or exploiting some symmetry. In this case the usual tangent half-angle substitution just introduces a detour in a straightforward solution.
Version 3. By considering the Fourier series of $\log\sin$ and $\log\cos$ we have that, in a distributional sense related to $L^2(-\pi/2,\pi/2)$,
$$ \cot\theta = 2 \sum_{k\geq 1} \sin(2k\theta) $$
$$ \log(\sin\theta+\cos\theta)=-\frac{\log 2}{2}-\sum_{k\geq 1}\frac{\cos(2k\theta+k\pi/2)}{k} $$
hence by Parseval's theorem
$$ \int_{0}^{\pi/2}\log(\sin\theta+\cos\theta)\cot(\theta)\,d\theta =\frac{\pi}{4}\sum_{\substack{k\geq 1\\k\text{ odd}}}\frac{(-1)^{(k-1)/2}}{k}=\frac{\pi}{4}\cdot\frac{\pi}{4} = \color{red}{\frac{\pi^2}{16}}\tag{6}$$
... WOW! This approach allows a simple and explicit evaluation of many integrals of the form $\int_{0}^{\pi/2}\log(\sin\theta+\cos\theta)\,\omega(\theta)\,d\theta$, thus many integrals of the form $\int_{0}^{1}\log(x+\sqrt{1-x^2})\,w(x)\,dx$. "Thinking backwards", the original problem can be probably tackled also by computing the moments $\int_{0}^{1}x^{2m+1} \log(x+\sqrt{1-x^2})\,dx$, then performing an interpolation/analytic continuation.