Test performance of parallel_flat_hash_map from https://github.com/greg7mdp/parallel-hashmap in 'pages' map in AsyncFileCached

A colleague of mine (Todd Lipcon) pointed out that we should test this performance improvement. He thought he witnessed a bottleneck in the "pages" map in AsyncFileCached.

Start by locating AsyncFileCached and its "pages" map, then read how that map is accessed and how performance tests are organized in the repository. Compare the existing map with parallel_flat_hash_map from the linked project and define a repeatable benchmark; done means the performance impact is measured and documented clearly.

Written by the indexing model from the issue text.