@gradio/dataframe@0.21.6

The Gradio team has released a targeted patch, version 0.21.6 for the `@gradio/dataframe` component, addressing a critical bug that could lead to silent data corruption. The issue, tracked as #12890, occurred when users sorted DataFrames within a Gradio interface, causing all NaN (Not a Number) values to be incorrectly replaced with zeros. This bug posed a significant risk for data science and machine learning demos, where preserving data integrity, especially for missing values represented as NaN, is paramount for accurate analysis and model interpretation.

The fix was implemented by community contributor @Mr-Neutr0n and merged by the automated `gradio-pr-bot`. This release also includes dependency updates for `@gradio/client` to version 2.1.0 and `@gradio/button` to 0.21.6. For developers, this underscores the importance of updating this specific component to maintain data fidelity in any application using interactive DataFrame displays. While a minor version bump, it is a crucial maintenance update for anyone relying on Gradio for data-centric AI applications, preventing a class of errors that could invalidate demo results without obvious warning.