refactor: simplify strategy UI and optimize Auto-K pipeline for 1,000–5,000+ notes

[Harsh16gupta]

This PR cleans up the webview UI and optimizes the clustering pipeline to scale smoothly for large note collections (500 to 5,000+ notes).

Key changes:

• UI cleanup: Removed raw silhouette scores from the webview and defaulted the view to K-Means. Added a short description card under the dropdown explaining HDBSCAN vs K-Means.
• Hides empty Uncategorized card: Automatically hides the empty Uncategorized card when K-Means is selected since K-Means categorizes 100% of notes.
• Dynamic Auto-K scaling: Auto-K range now scales dynamically for larger vaults (up to 2,000–5,000+ notes) so clusters stay balanced (~25–35 notes per category) without creating mega-clusters.
• UMAP & TF-IDF optimizations: Fixed intermediate UMAP to 5D, capped neighbors at 15, and cached TF-IDF n-grams for faster topic extraction.
• Version bump: Updated version to 0.1.6.

[Harsh16gupta]

The results which I get form the k-means algo are perfect; I like the clusters created, and how it scales to large note corpus
problem for now is that the hdbscan algo on a large corpus has a lot of notes that fall under the uncategorised category, so the next thing is to make it scale better. I will cater the hdbscan problem in the comming pr

for now we can merge this pr