ReplicaSetWriteBlockBypass::setFromMetadata is a write-blocking bypass initialization path, but it is currently executed even when the incoming operation is a read. The method sets the bypass state either from request metadata or from the AuthorizationSession default when the metadata field is absent, so calling it on reads introduces unnecessary state mutation / setup for operations that should not need replica-set write-block bypass handling.

As part of this ticket, we should:

1. Optimize readPrivilegedRequestMetadata so it does not call ReplicaSetWriteBlockBypass::get(opCtx).setFromMetadata(...) for read operations.
2. Remove the duplication in readPrivilegedRequestMetadata checks concerning global user write blocking and replica set write blocking.
3. Run a regression analysis for sep_benchmark and check that avoiding ReplicaSetWriteBlockBypass::setFromMetadata on reads reduces the extra read-path work introduced by SERVER-120970 and reported by BF-43448, with the expectation that instruction_per_iteration_mean for BM_FIND_ONE should go down.