Speed up JSON chunker.

[danielmai]

Remove a buffer size allocation in the JSON chunker that was taking up
a lot of time in the CPU profile while live loading, which had a huge
impact to live loading JSON data. I tried loading a JSON array of a
million objects structured like the following:

[{"uid":"0x1", "pred": "val"}, ...]

Before this change:

Number of TXs run            : 1010
Number of N-Quads processed  : 1009929
Time spent                   : 5m22.79806804s
N-Quads processed per second : 3136

After this change:

Number of TXs run            : 1010
Number of N-Quads processed  : 1009929
Time spent                   : 13.396047868s
N-Quads processed per second : 77686

Here's the CPU profile of the 5-minute live loading session 1-million edges (without this change), which shows the expensive GC path with (*jsonChunker).Chunk: cpu.pprof.gz.

go tool pprof -web ./cpu.pprof.gz

(pprof) list \).Chunk
Total: 11.83mins
ROUTINE ======================== github.com/dgraph-io/dgraph/chunker.(*jsonChunker).Chunk in src/github.com/dgraph-io/dgraph/chunker/chunk.go
     350ms   4.11mins (flat, cum) 34.72% of Total
         .          .    174:		return errors.Errorf("JSON file must contain array. Found: %v", ch)
         .          .    175:	}
         .          .    176:	return nil
         .          .    177:}
         .          .    178:
      20ms       20ms    179:func (*jsonChunker) Chunk(r *bufio.Reader) (*bytes.Buffer, error) {
      10ms      290ms    180:	out := new(bytes.Buffer)
         .   4.05mins    181:	out.Grow(1 << 20)
...

---

This change is 

[pullrequest]

✅ A review job has been created and sent to the PullRequest network.

---

Check the status or cancel PullRequest code review here.

[manishrjain]

LGTM. Please merge.