Add simple Langevin Example

I want to add a simple Langevin optimization example.

$$
\begin{aligned}
\text{Sample}: &&
\epsilon_k &\sim \mathcal{N}(0,1) \\
\text{Step}: &&
x_{k+1} &= x_k - \alpha\nabla_x f(x_k) + \sqrt{2\alpha}\epsilon_k
\end{aligned}
$$

Source:
* [blog](https://fa.bianp.net/blog/2023/ulaq/) | [blog](https://ericmjl.github.io/score-models/notebooks/02-langevin-dynamics.html) | [blog](https://www.jeremiecoullon.com/2020/11/10/mcmcjax3ways/) | [blog](https://bjlkeng.io/posts/bayesian-learning-via-stochastic-gradient-langevin-dynamics-and-bayes-by-backprop/)
* [Lecture Notes](https://www.di.ens.fr/appstat/spring-2022/lecture_notes/SGLD.pdf) |

No target file, notebook, entry point, or test is named. Start by locating existing optimization examples and notebook conventions in the repository, then use the proposed Langevin update and linked sources as references; done means a runnable simple example that demonstrates the stated sampling and step equations.

Written by the indexing model from the issue text.