Synchronous read of cluster files causes slow tasks

Processes in the cluster (server and client, I believe) read their cluster file at regular intervals of something like 5 or 10 minutes. Reading the cluster file is done synchronously using a call to [`readFileBytes`](https://github.com/apple/foundationdb/blob/e6d9ebd3d2c05509dc970768956c9c01d7a1c1e0/flow/Platform.actor.cpp#L2656). This call can sometimes take non-trivial amounts of time (100s of milliseconds), and the result is a slow task and increased latencies.

Start at flow/Platform.actor.cpp around the linked readFileBytes call, then trace the periodic cluster-file reads used by the server and client. Determine how the synchronous read creates slow tasks and increased latency; done means the cluster file can still be read at its regular intervals without blocking those tasks.

Written by the indexing model from the issue text.