Improve torrent search and ranking algorithm

Enhanced the ranking algorithm to better distinguish complete title matches from partial matches, reducing series confusion. Updated the torrent search modal to allow users to customize the search title for new requests, improving search flexibility. Also refined request lookup to ignore deleted requests.

src/lib/utils/ranking-algorithm.ts

@@ -201,10 +201,26 @@ export class RankingAlgorithm {
     // Title matching (0-35 points)
     let titleScore = 0;
     if (torrentTitle.includes(requestTitle)) {
-      // Exact substring match → full points
-      titleScore = 35;
+      // Found the title, but is it the complete title or part of a longer one?
+      const titleIndex = torrentTitle.indexOf(requestTitle);
+      const afterTitle = torrentTitle.substring(titleIndex + requestTitle.length);
+
+      // Title is complete if followed by clear metadata markers
+      // (not followed by more title words like "'s Secret" or " Is Watching")
+      const metadataMarkers = [' by ', ' - ', ' [', ' (', ' {', ' :', ','];
+      const isCompleteTitle = afterTitle === '' ||
+                              metadataMarkers.some(marker => afterTitle.startsWith(marker));
+
+      if (isCompleteTitle) {
+        // Complete title match → full points
+        titleScore = 35;
+      } else {
+        // Title continues with more words (e.g., "The Housemaid" + "'s Secret")
+        // This is likely a different book in a series → use fuzzy similarity
+        titleScore = compareTwoStrings(requestTitle, torrentTitle) * 35;
+      }
     } else {
-      // No exact match → use fuzzy similarity for partial credit
+      // No substring match at all → use fuzzy similarity
       titleScore = compareTwoStrings(requestTitle, torrentTitle) * 35;
     }