kv: don't load applied index twice during Raft snapshot

This was harmless, but it was also unnecessary and looked quite
alarming. It would be a major problem if the Raft applied index stored
in a `SnapshotMetadata` was ever incoherent with the applied state in
the snapshot. However, we're holding the raftMu in `snapshot`, so the
Raft applied index can't change. In fact, `StateLoader.Load` is only
atomic if we're holding the `raftMu` anyway. Still, better to avoid
questions.

The "members of the Range struct" comment was very stale, dating back
to acc41dd.

pkg/kv/kvserver/replica_raftstorage.go

@@ -575,21 +575,14 @@ func snapshot(
  return OutgoingSnapshot{}, errors.Mark(errors.Errorf("couldn't find range descriptor"), errMarkSnapshotError)
  }
 
- // Read the range metadata from the snapshot instead of the members
- // of the Range struct because they might be changed concurrently.
- appliedIndex, _, err := rsl.LoadAppliedIndex(ctx, snap)
+ state, err := rsl.Load(ctx, snap, &desc)
  if err != nil {
  return OutgoingSnapshot{}, err
  }
 
- term, err := term(ctx, rsl, snap, rangeID, eCache, appliedIndex)
- if err != nil {
- return OutgoingSnapshot{}, errors.Errorf("failed to fetch term of %d: %s", appliedIndex, err)
- }
-
- state, err := rsl.Load(ctx, snap, &desc)
+ term, err := term(ctx, rsl, snap, rangeID, eCache, state.RaftAppliedIndex)
  if err != nil {
- return OutgoingSnapshot{}, err
+ return OutgoingSnapshot{}, errors.Errorf("failed to fetch term of %d: %s", state.RaftAppliedIndex, err)
  }
 
  // Intentionally let this iterator and the snapshot escape so that the
@@ -606,7 +599,7 @@ func snapshot(
  RaftSnap: raftpb.Snapshot{
  Data: snapUUID.GetBytes(),
  Metadata: raftpb.SnapshotMetadata{
- Index: appliedIndex,
+ Index: state.RaftAppliedIndex,
  Term: term,
  // Synthesize our raftpb.ConfState from desc.
  ConfState: desc.Replicas().ConfState(),