To make this question easier, I'll put the center of the incircle in the origin. Any integral solution can be easily transferred to the location you indicated. Let $M,N,P$ be the corners of the triangle, as used in your question. Let the coordinates of $M$ be $(M_x,M_y)\in\mathbb Z^2$. One important quantity in your scenario is
$$d:=M_x^2+M_y^2-25$$
As you can see, as long as $M$ is outside the circle, that quantity will be positive. Its square root can be used to describe the tangents from $M$ to the circle:
\begin{align*}
t_1 &= \left\{(x,y)\;\middle\vert\;
(5M_x+\sqrt dM_y)x + (5M_y-\sqrt dM_x)y = 5\left(M_x^2+M_y^2\right) \right\} \\
t_2 &= \left\{(x,y)\;\middle\vert\;
(5M_x-\sqrt dM_y)x + (5M_y+\sqrt dM_x)y = 5\left(M_x^2+M_y^2\right) \right\}
\end{align*}
The way I obtained these tangents is by performing the dual of a conic-line intersection, as described in section 11.3 of Perspectives on Projective Geometry by Richter-Gebert.
If these tangents are to cross any points with integer coordinates, their slopes have to be rational (or $\infty$ for vertical lines). These slopes are
$$-\frac{5M_x\pm\sqrt dM_y}{5M_y\mp\sqrt dM_x}$$
There are two conceivable ways these slopes might be rational: either because $\sqrt d$ is rational, or because numerator and denominator are rational multiples of the same irrational number. If one treats $\mathbb Q[\sqrt d]$ as a two-dimensional $\mathbb Q$-vectorspace, then this translates into the requirement that the two vectors have to be linearily dependent, which here means
$$\det\begin{pmatrix}
5M_x & \pm M_y \\
5M_y & \mp M_x
\end{pmatrix}=\mp5\left(M_x^2+M_y^2\right)=0$$
So the only other way would mean $M$ is the origin, which is of course impossible. So we are left with the requirement that $\sqrt d$ be rational, so $d$ must be a square number.
Now one possible way to tackle this problem is iterating over all integral coordinates in a given range, and checking whether the resulting $d$ is square. One can build up a collection of possible corner points, and using the coordinates of the tangents, one can see which corners could belong to the same triangle. The $t_1$ of one point must match the $t_2$ of the other edge, but scalar multiples of the coefficients describe the same line. I've implemented such an enumeration. Many results contain $5$ as one of the coordinates of one of the points. But not all do, so one of the more interesting results is
\begin{align*}
M&=(-11, -2) & N&=( 1, 7) & P&=( 13, -9)
\end{align*}
After moving $C$ back to where you defined it, this becomes
\begin{align*}
M&=(-10, 0) & N&=( 2, 9) & P&=( 14, -7)
\end{align*}
However, the solution above has aright angle at $N$, so as your question asks for a slution without a right angle, this one is not acceptable. Thanks to @coffeemath for making me aware of this condition. It seems that at least if the absolute value of the coordinates is restricted to no more than $1000$, then all solution without a right angle neccessarily have either a horizontal or a vertical tangent. So here is another go, again first in the untranslated situation:
\begin{align*}
M&=(-11, -10) & N&=( -3, 5) & P&=( 25, 5)
\end{align*}
After moving $C$ back to where you defined it, this becomes
\begin{align*}
M&=(-10, -8) & N&=( -2, 7) & P&=( 26, 7)
\end{align*}
I also adapted my program to filter out the right-angled triangles automatically. There are still many possible solutions remaining, but none with all coordinates less than $25$ in absolute value, in the coordinate system where the circle center is the origin.
The side lengths are $a=25,b=39,c=56$ by the pythagorean theorem and the area is $\Delta=420$ by the Heron's formula or the shoelace formula, hence $r=7$. Since $h_c=\frac{2\Delta}{c}$, the length of the height relative to $C$ is $15$, hence the distance of $G$ from $AB$ is just $5$.
It is well known that $I=\frac{aA+bB+cC}{a+b+c}$ and $AI^2=bc-4rR=bc-\frac{2abc}{a+b+c}$, hence $IG^2$ can also be computed through the parallel axis theorem. Let we assume that $A$ has mass $a$, $B$ has mass $b$ and $C$ has mass $c$. With such assumptions, $I$ is the center of mass of $S=\{A,B,C\}$ and the moment of inertia of $S$ around $I$ is given by:
$$ M_I = a AI^2 + b BI^2 + c CI^2 = 3abc - 4(a+b+c) rR = abc $$
while the moment of inertia of $S$ around $G$ is given by:
$$ M_G = a AG^2 + b BG^2 + c CG^2 = \frac{1}{9}\sum_{cyc}a(2b^2+2c^2-a^2) $$
and by the parallel axis theorem $ M_G = (a+b+c) IG^2 + M_I$, hence:
$$ IG^2 = \frac{M_G-M_I}{a+b+c} = \boxed{\frac{\frac{2}{3}(a+b+c)^3-\frac{5}{3}(a^3+b^3+c^3)-13abc}{9(a+b+c)}}.$$
A difficult problem has just born:
Prove that if $a,b,c$ are the side lengths of a triangle,
$$ 2(a+b+c)^3 \geq 5 (a^3+b^3+c^3) + 39abc. $$
Tricky solution: $IG^2\geq 0$, and equality is achieved only when $I\equiv G$, i.e. when $a=b=c$.
Best Answer
The triangle is only determined up to rotations and reflections.
Let $G$ be the centroid of $\triangle ABC$. The length of the medians can be calculated in terms of the sides using the median length formula. That gives the length of $GA$ as $2/3$ the length of the median through $A$. Choose $A$ to be any point on the circle centered at $G$ of radius $GA$. Now construct $B$ knowing the distances $AB,GB\,$, then $C$ the same way. Depending on the choice of $B$ on either side of $GA$ this gives two possible triangles $ABC$ which are reflections of each other.