Like uhoh said in a comment, while $t(r)$ is analytical, $r(t)$ to my knowledge is not.
$t(r)$ can however be used very efficiently to approximate $r(t)$ since every iteration of $t(r)$ will give you the same number of extra digits, thus converging very quickly.
Here is the derivation of an analytical $t(r)$.
- The distance from $r_P$ to $K$ (the Earth projected down on the apsis line), let's call it $d$, is:
$$d = \frac{2a(r - r_P)}{(r_A - r_P)}$$
- The angle between the apsis line and the line from the geometric centre to the Earth projected out on the semi-major circle is:
$$\beta =\cos^{-1}\left(\frac{a-d}{a}\right)$$
- The projected swept area is now the sector area minus the triangle between the Sun, the geometric centre and the projected Earth.
$$A_{proj} = \frac{\beta a^2}{2} - \frac{(a - r_P)\sqrt{a^2 - (a - d)^2}}{2}$$
- From that, we can get the real swept area by scaling back by the ratio between the semi-major and semi-minor axis:
$$A_{swept} = A_{proj} \cdot \frac{b}{a}$$
- Which from Kepler's third law can be used to calculate the time since periapsis:
$$T_{since\ periapsis} = 2\pi\sqrt{\frac{a^3}{\mu}} \cdot \frac{A_{swept}}{ab\pi}$$