fix: use char boundaries instead of byte indices for UTF-8 strings in notifications

PANIC FIX: Notification preview truncation was using byte indices (`[..147]`)
instead of char boundaries, causing panic when truncating UTF-8 strings
containing multi-byte characters (Cyrillic, emoji, etc.).

Error message:
"byte index 147 is not a char boundary; it is inside 'п' (bytes 146..148)"

Fix:
- Replace `beautified.len() > 150` with `beautified.chars().count() > MAX_PREVIEW_CHARS`
- Replace `&beautified[..147]` with `beautified.chars().take(MAX_PREVIEW_CHARS).collect()`
- Add constant MAX_PREVIEW_CHARS = 147 for clarity

This ensures we truncate at character boundaries, not byte boundaries,
preventing panics on multi-byte UTF-8 sequences.

Impact: Notifications for messages with Russian/emoji text crashed the app.

Root cause: Classic Rust UTF-8 indexing mistake - slicing by bytes instead
of characters.

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

src/notifications.rs

@@ -164,9 +164,12 @@ impl NotificationManager {
                 // Beautify media labels with emojis
                 let beautified = Self::beautify_media_labels(text);
 
-                // Limit preview length
-                if beautified.len() > 150 {
-                    format!("{}...", &beautified[..147])
+                // Limit preview length (use char count, not byte count for UTF-8 safety)
+                const MAX_PREVIEW_CHARS: usize = 147;
+                let char_count = beautified.chars().count();
+                if char_count > MAX_PREVIEW_CHARS {
+                    let truncated: String = beautified.chars().take(MAX_PREVIEW_CHARS).collect();
+                    format!("{}...", truncated)
                 } else {
                     beautified
                 }