Add stats and metrics reprs

[c24t]

This PR adds __repr__ methods for a few stats and metrics classes in the style of @reyang's changes in #259.

>>> metric
Metric(descriptor.name="sum by oslang")

>>> metric.descriptor
MetricDescriptor(name="sum by oslang", description="Sum of something by os, lang", unit=kW/s, type=ValueDouble)

>>> metric.time_series[0]
TimeSeries(points=[Point(value=ValueDouble(12.34), timestamp=2019-01-31 11:28:58.553910)], label_values=[LabelValue(android), LabelValue(en_us)], start_timestamp=2019-01-31T19:28:38.383422Z)

>>> metric.time_series[0].points[0]
Point(value=ValueDouble(12.34), timestamp=2019-01-31 11:28:58.553910)

>>> vd.view.aggregation.aggregation_data
SumAggregationDataFloat(0.0)

>>> metric.time_series[0].points[0]
Point(value=ValueDouble(12.34), timestamp=2019-01-31 11:28:58.553910)

Not all reprs include all the attributes of the class, and in general don't allow the user to reconstruct the object by evaling it. I'd prefer for reprs to be short and human-readable, and not an exhaustive description of the object, but this is potentially confusing if the user doesn't realize that not every attribute is reflected in the repr. I'm interested to hear other reviewers' opinions.

This PR doesn't include reprs for all classes, just ones that were helpful while writing #476. If you agree with this style we can add reprs to new objects going forward.