Clarify synthetic string benchmark purpose

Author: lmeyerov

benchmarks/gfql/filter_pagerank/benchmark_to_cugraph_gpu.py

@@ -7,6 +7,7 @@
 - whether `store_transposed=True` helps enough to keep locally
 - whether `renumber=False` is semantically safe on a given graph family
 - whether CUDA / RAPIDS image changes affect graph-build vs pagerank time separately
+- whether repeated low-cardinality string ids reproduce the `gplus`-style renumber gap
 """
 from __future__ import annotations