Actually use the current time range for querying sections

Doh!
Without this, we keep inferring modes for the same sections over and over.

Analysis below.

------------------------------------------------------------------------

Aha! we now have two copies of the analysis results

```
In [19]: edb.get_analysis_timeseries_db().find({"metadata.key": "inference/prediction"},
    ...:  {"data.predicted_mode_map": "1"}).count()
Out[19]: 137

In [20]: edb.get_analysis_timeseries_db().find({"metadata.key": "analysis/inferred_secti
    ...: on"}, {"data.sensed_mode": "1"}).count()
Out[20]: 137

In [21]: edb.get_analysis_timeseries_db().find({"metadata.key": "analysis/inferred_secti
    ...: on"}, {"data.sensed_mode": "1"}).count()
Out[21]: 274

In [22]: edb.get_analysis_timeseries_db().find({"metadata.key": "inference/prediction"},
    ...:  {"data.predicted_mode_map": "1"}).count()
Out[22]: 274
```

Let's figure out why?!

Ok so when we get to the MODE_INFERENCE stage, the `start_ts` is set correctly.

```
2018-02-26 10:03:41,381:INFO:140735691387712:For stage PipelineStages.MODE_INFERENCE, start_ts = 2016-10-08T01:34:35.600000
```

But we still read all the sections again. Ah, the query doesn't appear to
involve the timestamp at all.

```
2018-02-26 10:03:41,383:DEBUG:140735691387712:curr_query = {'user_id': UUID('ea07ac60-0724-4eec-aab2-8240fcbf74ce'), '$or': [{'metadata.key': 'analysis/cleaned_section'}]}, sort_key = metadata.write_ts
```

emission/analysis/classification/inference/mode/pipeline.py

@@ -126,7 +126,7 @@ def getLastSectionDone(self):
   def runPredictionPipeline(self, user_id, timerange):
     self.ts = esta.TimeSeries.get_time_series(user_id)
     self.toPredictSections = esda.get_entries(esda.CLEANED_SECTION_KEY, user_id, 
-        time_query=None)
+        time_query=timerange)
     if (len(self.toPredictSections) == 0):
         logging.debug("len(toPredictSections) == 0, early return")
         assert self.last_section_done is None, ("self.last_section_done == %s, expecting None" % \