Fix BTree cache CPU trashing eviction

Cache refuses to evict inner nodes, and scans the entire cache looking for non-inner nodes, which become increasingly rare.

Ditch the inner node policy, instead just ensure LRU is updated on get(). Why was it not already doing so ? That is the point of an LRU.

I was seeing queries where > 70% of the CPU was spend on BTreeCache.removeNext(). That’s completely fixed; such queries are now 5x faster.

exist-core/src/main/java/org/exist/util/hashtable/SequencedLongHashMap.java

@@ -75,7 +75,7 @@ public SequencedLongHashMap(final int iSize) {
      * @return the value associated with the key, or null if the key is absent
      */
     public @Nullable V get(final long key) {
-        return map.get(key);
+        return map.getAndMoveToLast(key);
     }
 
     /**