Consider $(1)$
How does one verify that
$$\int_{0}^{1}x\ln[\color{red}{\ln{x}}\ln(1-x)]\mathrm dx=-\gamma\tag1$$
An attempt:
$$\int_{0}^{1}x\ln{(\ln{x})}\mathrm dx+\int_{0}^{1}x\ln{(\ln{(1-x))}}\mathrm dx\tag2$$
$u=\ln{x}$ then $xdu=dx$
$$\int_{0}^{\infty}e^{-2u}\ln{(2u)}\mathrm du+\int_{0}^{\infty}e^{-2u}\ln{(1-e^u)}\mathrm du=I_1+I_2\tag3$$
Applying integration by parts to $I_1$
$$I_1={-e^{-2u}\ln{(2u)}\over 2}|_{0}^{\infty}+{1\over 2}\int_{0}^{\infty}{1\over u}\cdot e^{-2u}\mathrm du\tag4$$
$(4)$ seem to be divergent
How else can we tackle $(1)?$
There's a subtle problem with your split: the terms you've defined, e.g. $x\ln(\ln(x))$, aren't actually defined (over the reals) on the domain $(0,1)$ because $\ln(x)$ is negative on this range! Instead, you have to split things as $\displaystyle\int_0^1x\ln(-\ln(x))dx+\int_0^1x\ln(-\ln(1-x))dx$; this works because the minus signs internally cancel when the terms are multiplied!
But now we can substitute $u=1-x$ into the second term: minus signs from $du=-dx$ and from switching the order of integration from $\int_1^0$ back to $\int_0^1$ cancel, leaving the second integral as $\displaystyle\int_0^1(1-u)\ln(-\ln(u))du$. And then we can add this to the first term of the split to get $\displaystyle\int_0^1\ln(-\ln(x))dx$. But this formula (more often written in the form $\displaystyle\int_0^1\ln(\ln(\frac1x))dx$) is one of the classical expressions for $-\gamma$.