Persistence rate-limit ResourceExhausted is flattened to Unavailable in ProcessOutgoingSearchAttributes

Actual behaviour

service/history/api/get_history_util.go:358 converts a *serviceerror.ResourceExhausted from the
persistence rate limiter into a *serviceerror.Unavailable:

saTypeMap, err := saProvider.GetSearchAttributes(persistenceVisibilityMgr.GetIndexName(), false)
if err != nil {
      return serviceerror.NewUnavailablef(consts.ErrUnableToGetSearchAttributesMessage, err)
}

Expected behaviour

ErrPersistenceSystemLimitExceeded reaches the caller as ResourceExhausted, so the gRPC status reflects rate limiting and common/backoff/retry.go:80 applies the throttle backoff.

Impact

• Wrong status code. Rate limiting is reported as Unavailable.
• Throttle retry bypassed. common/backoff/retry.go:80 applies the longer backoff only to
  *serviceerror.ResourceExhausted, so retries run at the ordinary faster rate against an already exhausted persistence budget. This contradicts the intent stated at service/history/queues/executable.go:795.

Specifications

• Version: 1.31.2
• Platform: self-hosted postgres

Read service/history/api/get_history_util.go:358, then trace the error handling through common/backoff/retry.go:80 and service/history/queues/executable.go:795. Confirm the persistence rate-limit error remains ResourceExhausted to the caller and that the throttle backoff path is used, with regression coverage if the surrounding tests identify a suitable location.