Cherry pick changes from #15138; add MIGRATE.md entry

Author: uschindler

lucene/CHANGES.txt

@@ -270,6 +270,11 @@ Optimizations
 
 * GITHUB#15045: Use FixedBitSet#cardinality for counting liveDocs in CheckIndex (Zhang Chao)
 
+* GITHUB#15116, GITHUB#15138: Rewrite of the GroupVInt optimization without lambdas, varhandles
+  and no code in subclasses. The optimization now auto-detects if an IndexInput supports random access
+  and uses an optimized branchless approach. Any subclasses that have implemented the optimized method
+  need to remove it as it will disappear in Lucene 11.  (Uwe Schindler)
+
 Changes in Runtime Behavior
 ---------------------
 * GITHUB#14823: Decrease TieredMergePolicy's default number of segments per

lucene/MIGRATE.md

@@ -35,6 +35,13 @@ long maxRamBytesUsed = 50 * 1024 * 1024; // 50MB
 IndexSearcher.setDefaultQueryCache(new LRUQueryCache(maxCachedQueries, maxRamBytesUsed));
 ```
 
+### Possibility to optimize `readGroupVInt()` in `DataInput` subclasses removed (GITHUB#15116)
+
+Any subclass of `DataInput` that have implemented `readGroupVInt()` need to remove that implementation.
+
+Instead make sure that subclasses of `IndexInput` implement `RandomAccessInput`.
+Pure `DataInput` subclasses cannot be optimized anymore as they cannot offer random access and seeking.`
+
 ## Migration from Lucene 9.x to Lucene 10.0
 
 ### DataInput#readVLong() may now read negative vlongs