Optimize Stored Fields Processing by Avoiding Redundant Loop for Checking Array Element Equality

### Description
I noticed that when Lucene processes stored fields, it saves `numStoredFields` and `lengths` as shown in the code below:

```java
// save numStoredFields
saveInts(numStoredFields, numBufferedDocs, fieldsStream);

// save lengths
saveInts(lengths, numBufferedDocs, fieldsStream);

private static void saveInts(int[] values, int length, DataOutput out) throws IOException {
if (length == 1) {
out.writeVInt(values[0]);
} else {
StoredFieldsInts.writeInts(values, 0, length, out);
}
}
```
Source: [Lucene90CompressingStoredFieldsWriter.java](https://github.com/apache/lucene/blob/caeabf39309a91997d361b4104bda105d16ae720/lucene/core/src/java/org/apache/lucene/codecs/lucene90/compressing/Lucene90CompressingStoredFieldsWriter.java#L217-L221)

During the `StoredFieldsInts.writeInts(values, 0, length, out);` operation, the method iterates through the `values` array to check if all the elements are the same. I propose an optimization to potentially increase performance by avoiding this loop.

My suggestion is to add two new fields: `numStoredFieldsAllSame` and `docLengthAllSame`. These fields would track whether all the elements in the respective arrays are the same during insertion. By passing these fields to the `writeInts` method, we can eliminate the need for the loop.

Start with Lucene90CompressingStoredFieldsWriter.java, especially saveInts and the insertion paths for numStoredFields and lengths. Then inspect StoredFieldsInts.writeInts to understand its array-equality check and how the proposed flags would affect encoding. Done means the flags are maintained, passed into writeInts, and the redundant equality scan is avoided without changing stored-field output.

Written by the indexing model from the issue text.