Fix remaning jepsen issues

[janardhan1993]

This change is 

[janardhan1993]

Will remove linearizability from this PR. Kept it for testing

---

Review status: 0 of 6 files reviewed at latest revision, all discussions resolved.

---

dgraph/cmd/zero/raft.go, line 574 at r2 (raw file):

					n.server.updateLeases()
				}
				leader = rd.RaftState == raft.StateLeader

There was a bug here, we weren't setting leader to false when it steps down. So once a node steps down and then becomes leader again the lease was continuing from the leader was leader. (So readTs was going back)

---

posting/list.go, line 348 at r1 (raw file):

		l.markdeleteAll > 0 && t.Op == intern.DirectedEdge_DEL &&
		bytes.Equal(t.Value, []byte(x.Star))
	doAbort := hasPendingDelete || txn.StartTs < l.commitTs

This needn't be true always for index keys, leave conflict detection to zero shouldn't do here.

Example:
<0x1> "abc" . say startts 1 committs 2
<0x2> "abc" , startts 3 committs 4
now for index key "abc" anything can get applied first
on replica
because scheduler does dependency checking only at sp level
even if we don't do here zero would do conflict detection
Only difference is user would get error while committing instead of mutating

---

worker/draft.go, line 125 at r1 (raw file):

var errReadIndex = x.Errorf("cannot get linerized read (time expired or no configured leader)")

func (n *node) WaitLinearizableRead(ctx context.Context) error {

Copied linerizability implementation from zero/raft.go to check whether it fixes the issue or not.
Will change it to batching implementation

---

Comments from Reviewable

[pawanrawal]

some nice fixes there.

---

Reviewed 1 of 5 files at r1, 1 of 2 files at r2, 3 of 4 files at r3.
Review status: all files reviewed at latest revision, 3 unresolved discussions.

---

dgraph/cmd/zero/raft.go, line 574 at r2 (raw file):

Previously, janardhan1993 (Janardhan Reddy) wrote…

There was a bug here, we weren't setting leader to false when it steps down. So once a node steps down and then becomes leader again the lease was continuing from the leader was leader. (So readTs was going back)

good one.

---

worker/mutation.go, line 416 at r3 (raw file):

	if txn := posting.Txns().Get(startTs); txn != nil {
		txn.Fill(tctx)
		if l := len(txn.Indices); l > 0 {

Can we use txn.LastIndex() here? Also, can that function have a RLock instead of Lock?

---

Comments from Reviewable

[janardhan1993]

Review status: 4 of 5 files reviewed at latest revision, 4 unresolved discussions.

---

worker/mutation.go, line 416 at r3 (raw file):

Previously, pawanrawal (Pawan Rawal) wrote…

Can we use txn.LastIndex() here? Also, can that function have a RLock instead of Lock?

Done.

---

Comments from Reviewable