Make results more stable by including phrase in sorting.

Pre-heat cache:
DID 4017 on 1 threads in 4.944, 812.51/s (812.5 per thread), 1.2308ms/q (4014/4017)
Cache Cities/Streets {'hits': 1100, 'misses': 2917, 'inserts': 2917, 'size': 2000} {'hits': 1369, 'misses': 1305, 'inserts': 1305, 'size': 1305}
OK:
DID 4017 on 1 threads in 4.328, 928.24/s (928.2 per thread), 1.0773ms/q (4014/4017)
DID 4017 on 4 threads in 1.501, 2677.06/s (669.3 per thread), 0.3735ms/q (4014/4017)
DID 4017 on 8 threads in 1.039, 3867.01/s (483.4 per thread), 0.2586ms/q (4014/4017)
DID 4017 on 12 threads in 0.930, 4321.07/s (360.1 per thread), 0.2314ms/q (4014/4017)
DID 4017 on 16 threads in 0.894, 4493.22/s (280.8 per thread), 0.2226ms/q (4014/4017)
(All recent tests on AMD Ryzen 7 3800X 8-Core, 32GB RAM)

Author: Tomasz bla Fortuna

src/fuzzdex/seeker.rs

@@ -246,8 +246,8 @@ impl Index {
         }
 
         results.sort_unstable_by(|a, b| {
-            let side_a = (a.distance, -a.score, -a.should_score, a.origin.len());
-            let side_b = (b.distance, -b.score, -b.should_score, b.origin.len());
+            let side_a = (a.distance, -a.score, -a.should_score, a.origin.len(), &a.origin);
+            let side_b = (b.distance, -b.score, -b.should_score, b.origin.len(), &b.origin);
             side_a.partial_cmp(&side_b).unwrap_or(Ordering::Equal)
         });