Should we enable BP reordering for the HNSW graph?

[mikemccand]

This exciting Lucene change will reorder Lucene documents to create a more compact HNSW graph while merging.

It's useful for vector-heavy Lucene indices that do not otherwise have a preference on how the documents are statically sorted.

The nightly benchy now tests KNN indexing+searching in several places: twice from nightlyBench.py for quantized and unquantized highly concurrent indexing performance, again in nightlyBench.py into the single-threaded consistent-segment-geometry search index, and again in runNightlyKnn.py which runs knnPerfTest in several configurations.

Let's try turning on this new BP reordering in some of these?