Here is a path for getting
$$
\sum_{\substack{0\le n\\1\le k\le(n+1)}}
\frac 1k\cdot H_{k-1/2}\; x^n
=
\frac{2}{x(1-x)}
\Big( \operatorname{arctanh}^2 \sqrt{x}
+
\log 2 \log (1-x) \Big)\ .
$$
(We show this holds as an equality of meromorphic functions in the open
unit disk in the complex plane.)
Using $$H_{k-1/2}=(2H_{2k}-H_k)-2\log 2\ ,$$the relation to be shown splits in two parts, we consider them separately. The simpler part is the one with coefficients (in each $x^n$) in $2\log 2\cdot\Bbb Q$, we consider it first, it is a good warm-up. The slightly more complicated part with rationaly coefficients (in $x^n$) is similar, we need to write the Cauchy product of two series, one being again $1/(1-x)=(1+x+x^2+x^3+\dots)$, with more care.
The simpler series, we substitute $N=n+1\ge1$ explicitly, to make it simpler for the bare eye to follow at one glance:
$$
\begin{aligned}
\sum_{\substack{0\le n\\1\le k\le(n+1)}}
\frac 1k\; x^n
&=
\frac 1x
\sum_{\substack{1\le N\\1\le k\le N}}
\frac 1k\; x^N
\\
&=\frac 1x\sum_{1\le N}\left(\frac 11+\frac 12+\dots+\frac 1N\right)x^N
\\
&=\frac 1x
(1+x+\dots+x^N+\dots)
\left(\frac x1+\frac {x^2}2+\dots+\frac {x^N}N+\dots\right)
\\
&=\frac 1x
\cdot\frac 1{1-x}
\cdot\Big(\ -\ln(1-x)\ \Big)\ .
\end{aligned}
$$
The other part is similar. We isolate the same type of a Cauchy product, one of the factors being $1/(1-x)$. After multiplication with $x$:
$$
\begin{aligned}
x\sum_{\substack{0\le n\\1\le k\le(n+1)}}
\frac 1k(2H_{2k}-H_k)\; x^n
&=
\sum_{\substack{1\le N\\1\le k\le N}}
\frac 1k(2H_{2k}-H_k)\; x^N
\\
&\qquad\text{ and we keep this in mind, on the other side,}
\\
&\qquad\text{ using $y=\sqrt x$:}
\\[3mm]
x\cdot\frac{2}{x(1-x)}
\operatorname{arctanh}^2 \sqrt{x}
&=
\frac 2{1-y^2}\left(\sum_{j\ge 1\text{ odd}}\frac 1j y^j\right)^2
\\
&=
\frac 2{1-y^2}\sum_{2\le M\text{ even}}
\sum_{\substack{
j\ge 1\text{ odd}\\
k\ge 1\text{ odd}\\
j+k = M
}}\frac 1j\cdot \frac 1k \; y^M
\\
&=
2\Big(1+y^2+y^4+y^6+\dots\Big)\sum_{2\le M\text{ even}}
\sum_{\substack{
j\ge 1\text{ odd}\\
k\ge 1\text{ odd}\\
j+k = M
}}\frac 1j\cdot \frac 1k \; y^M
\\
&=
2
\sum_{2\le M\text{ even}}
\sum_{\substack{
j\ge 1\text{ odd}\\
k\ge 1\text{ odd}\\
j+k \le M
}}
\frac 1j\cdot \frac 1k \; y^M
\text{ and with $M=2N$, $x=y^2$,}
\\
&=
2
\sum_{1\le N\text{ arbitrary}}
\sum_{\substack{
j\ge 1\text{ odd}\\
k\ge 1\text{ odd}\\
j+k \le 2N
}}
\frac 1j\cdot \frac 1k \; x^N
\ .
\end{aligned}
$$
Let us try now to figure out the simplification scheme for the two expressions of the coefficients in $x^N$, $N$ integer $\ge 1$, for a fixed small value of $N$. We take $N=5$. The sum over the $\frac 1j\cdot \frac 1k$ with odd $j,k\ge 1$, having sum $\le 2N$ has the terms explicitly:
$$
\text{twice}\qquad
\begin{array}{ccccc}
\frac 11\cdot \frac 11 &
\frac 11\cdot \frac 13 &
\frac 11\cdot \frac 15 &
\frac 11\cdot \frac 17 &
\color{magenta}{\frac 11\cdot \frac 19 }
\\
\frac 13\cdot \frac 11 &
\frac 13\cdot \frac 13 &
\frac 13\cdot \frac 15 &
\color{magenta}{\frac 13\cdot \frac 17} &
\\
\frac 15\cdot \frac 11 &
\frac 15\cdot \frac 13 &
\color{magenta}{\frac 15\cdot \frac 15} &
&
\\
\frac 17\cdot \frac 11 &
\color{magenta}{\frac 17\cdot \frac 13} &
&
&
\\
\color{magenta}{\frac 19\cdot \frac 11} &
&
&
&
\end{array}
$$
The other side is
$$
\begin{aligned}
\sum_{1\le k\le N}
\frac 1k(2H_{2k}-H_k)
=&
\frac 11\left(\frac 21\color{blue}{+\frac 22-\frac 11}\right)
\\
+&
\frac 12\left(\frac 21\color{blue}{+\frac 22}+\frac 23\color{blue}{+\frac 24-\frac 11-\frac 12}\right)
\\
+&
\frac 13\left(\frac 21\color{blue}{+\frac 22}+\frac 23\color{blue}{+\frac 24}+\frac 25\color{blue}{+\frac 26-\frac 11-\frac 12-\frac 13}\right)
\\
+&
\frac 14\left(\frac 21\color{blue}{+\frac 22}+\frac 23\color{blue}{+\frac 24}+\frac 25\color{blue}{+\frac 26}+\frac 27\color{blue}{+\frac 28-\frac 11-\frac 12-\frac 13-\frac 14}\right)
\\
+&
\frac 14\left(\frac 21\color{blue}{+\frac 22}+\frac 23\color{blue}{+\frac 24}+\frac 25\color{blue}{+\frac 26}+\frac 27\color{blue}{+\frac 28}+\frac 29\color{blue}{+\frac 2{10}-\frac 11-\frac 12-\frac 13-\frac 14-\frac 15}\right)
\\[3mm]
=&
\frac 11\left(\frac 21\right)
\\
+&
\frac 12\left(\frac 21+\frac 23\right)
\\
+&
\frac 13\left(\frac 21+\frac 23+\frac 25\right)
\\
+&
\frac 14\left(\frac 21+\frac 23+\frac 25+\frac 27\right)
\\
+&
\color{magenta}{
\frac 15\left(\frac 21+\frac 23+\frac 25+\frac 27+\frac 29\right)}
\\[3mm]
\ .
\end{aligned}
$$
So inductively we would have to extrapolate the role of the magenta terms, then show their equality inductively, explicitly:
$$
\color{magenta}{
\sum_{\substack{1\le j,k\text{ odd}\le 2N\\j+k=2N}}
\frac 1j\cdot\frac 1k
=
\frac 1N\sum_{1\le j\text{ odd}\le 2N}\frac 1j
\ .}
$$
Showing this inductively would break the harmonic beauty, so let us do the job thematically. And this again in the special case $N=5$, since the idea can be easily extrapolated. (It is a typical MZV (multiple zeta values) idea.) The integral form for the equality to be shown is (with $a,b$ playing the roles of $j-1$, $k-1$):
$$
\sum_{\substack{0\le a,b\text{ even}\le 2N-2\\a+b=2N-2}}
\iint_{[0,1]^2}x^a\; y^b\; dx\; dy
=
\frac 2{2N}
\sum_{0\le a\text{ even}\le 2N-2}t^a\; dt
\ .
$$
The idea is now to write the double integral on the square as twice the double integral on the simplex $0\le x\le y\le 1$, then change variables, $y=Y$, $x=tY$ with $t,Y\in[0,1]$. Then (formally) $dx\wedge dy = d(tY)\wedge dY=Y\; dt\wedge dY$, and we get a uniform factor $Y^{a+b+1}=Y^{2N-1}$, and by integration the factor $\frac1{2N}$.
We have now completed also the second part.
This became a long answer, but i hope the combinatorial idea behind the steps became in this way easy and transparent.