Replace separate logo images with combined logos

Author: GermanHeim

paper/figures/logo-py.png

@@ -45,9 +45,9 @@ $$
 $$
 
 Developed in Rust—a free and open-source programming language—it combines accessibility with performance, ensuring that researchers are not limited by proprietary ecosystems.
-Rust’s ownership model and concurrency features enable safe and efficient execution, making the library suitable for high-performance computing applications, a suitability further supported by recent work demonstrating Rust’s advantages in high-performance computing [@eth-rust].
+Rust's ownership model and concurrency features enable safe and efficient execution, making the library suitable for high-performance computing applications, a suitability further supported by recent work demonstrating Rust's advantages in high-performance computing [@eth-rust].
 
-![`GlobalSearch-rs` logo.\label{fig:logo-globalsearch-rs}](./figures/logo-rs.png){width=75%}
+![`GlobalSearch-rs` (right) and `PyGlobalSearch` (left) logos.\label{fig:logos}](./figures/logos.png)
 
 # Library features
 
@@ -59,8 +59,6 @@ Parallel execution is a core feature of `GlobalSearch-rs`, utilizing the `Rayon`
 
 As an open-source and cross-platform library, `GlobalSearch-rs` is designed to be accessible to researchers and engineers from various fields. The Python bindings, created using `PyO3` and `maturin`, enable users to leverage its capabilities within Python-based numerical and data science environments (e.g., `NumPy` [@numpy], `JAX` [@jax]), making it a practical choice for interdisciplinary research and engineering applications.
 
-![`PyGlobalSearch` logo.\label{fig:logo-pyglobalsearch}](./figures/logo-py.png){width=75%}
-
 The software is distributed under the permissive MIT license, encouraging use, modification, and contribution from the community. To ensure reliability and maintain quality, continuous integration testing is performed via GitHub Actions, providing automated validation of builds and functionality across supported platforms.
 
 # Example: Optimizing the Six-Hump Camel Function