Lag Reported with no lag in any interval

Nobody has claimed this yet.

A lag of `1` is being reported over various channels (HipChat, PagerDuty), but when drilling down into the partition, no interval is reporting lag. How can this be prevented?

```
{
"offsets": [
{
"offset": 1956316,
"timestamp": 1542361198191,
"lag": 0
},
{
"offset": 1956316,
"timestamp": 1542361318612,
"lag": 0
},
{
"offset": 1956316,
"timestamp": 1542361437805,
"lag": 0
},
{
"offset": 1956316,
"timestamp": 1542361558106,
"lag": 0
},
{
"offset": 1956316,
"timestamp": 1542361678486,
"lag": 0
},
{
"offset": 1956316,
"timestamp": 1542361797643,
"lag": 0
},
{
"offset": 1956316,
"timestamp": 1542361918593,
"lag": 0
},
{
"offset": 1956316,
"timestamp": 1542362037623,
"lag": 0
},
{
"offset": 1956316,
"timestamp": 1542362158668,
"lag": 0
},
{
"offset": 1956316,
"timestamp": 1542362277504,
"lag": 0
},
{
"offset": 1956316,
"timestamp": 1542362397963,
"lag": 0
},
{
"offset": 1956316,
"timestamp": 1542362518529,
"lag": 0
},
{
"offset": 1956316,
"timestamp": 1542362637295,
"lag": 0
},
{
"offset": 1956316,
"timestamp": 1542362758359,
"lag": 0
},
{
"offset": 1956316,
"timestamp": 1542362878464,
"lag": 0
}
],
"owner": "",
"current-lag": 1
},
```

Start by reproducing the reported alert using the offsets JSON, comparing the zero interval lag values with the reported current-lag of 1. Trace how Burrow calculates and sends lag notifications to HipChat and PagerDuty. Done means identifying the discrepancy, preventing false lag alerts, and covering the case with a regression test.

Written by the indexing model from the issue text.