Mutex bottleneck upon reading from database

This is using 5.14.3 on Windows, I believe I am only able to read about 15MB/s from the database or so. When I profiled rocksdb, it appears that resizing of a vector when reading data in is causing a big bottleneck. It is unclear if there is a setting to fix this value, as making the block size bigger did not work.

No source file, test, or entry point is named. Start by reproducing the Windows read-throughput issue on RocksDB 5.14.3 and reviewing the supplied profile for vector resizing during reads; compare results with different block-size settings. Done means the bottleneck is identified and a verified change or configuration resolves the reported throughput problem.

Written by the indexing model from the issue text.