Support disabling conflict detection

[jarifibrahim]

This commit adds support for disabling conflict detection by setting the
option.DetectConflicts=false. When conflict detection is disabled
badger will not store information required to detect the conflict. This
reduces the amount of memory used by transactions running parallely and
also allows transactions to be processed at a faster rate.

fixes BADGER-207

---

This change is 

[martinmr]

Reviewed 2 of 2 files at r1.
Reviewable status: all files reviewed, 1 unresolved discussion (waiting on @ashish-goswami, @jarifibrahim, and @manishrjain)

---

txn.go, line 654 at r1 (raw file):

// tree won't be updated, so there's no need for any rollback.
func (txn *Txn) Commit() error {
	if len(txn.pendingWrites) == 0 {

why did this change?

[manishrjain]

A bunch of comments.

Reviewed 2 of 2 files at r1.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @ashish-goswami, @jarifibrahim, and @martinmr)

---

txn.go, line 391 at r1 (raw file):

	if txn.db.opt.DetectConflicts {
		fp := z.MemHash(e.Key) // Avoid dealing with byte arrays.
		txn.writes[fp] = struct{}{}

In Dgraph, we call them conflict keys or something. Maybe call it the same. Don't call it writes, then it's hard to read.

---

txn.go, line 654 at r1 (raw file):

Previously, martinmr (Martin Martinez Rivera) wrote…

why did this change?

Just add a comment. That writes might be empty if the conflict detection is disabled. Therefore, we should check the pending writes.