support per-org metrics_active for scraping by prometheus

[woodsaj]

closes #1159

[Dieterbe]

to be clear, we have 3 places where we track active metrics:

1. idx.metrics_active (carbon)
2. tank.metrics_active (carbon)
3. metrics_active (prometheus) ("tank" implied)

only 3 is changing, not 1/2. so please update PR title and/or description

Also, we should get a sense for any performance impact of this change.
for the ingest path:
BenchmarkProcessMetricDataSameMetric
BenchmarkProcessMetricDataUniqueMetrics
these exercise AggMetrics.GetOrCreate(). we currently don't have benchmarks for AggMetrics.Get() (which is exercised for render/getData queries) but I imagine the Get() call only plays a very minor role in it, same for the role of GetOrCreate in the benchmarks mentioned above, but why not measure them anyway to be sure.

we don't have benchmarks for AggMetrics.GC(). the changes would only play a minor role and we don't really care how long the operation runs, as long as the write locks aren't worse, which is the case here.

[woodsaj]

no noticeable change in benchmark performance. The +/- 2% is due to my desktop doing other things.

$ benchcmp master.txt perOrgActiveMetrics.txt
benchmark                                     old ns/op     new ns/op     delta
BenchmarkProcessMetricDataUniqueMetrics-4     35.1          34.4          -1.99%
BenchmarkProcessMetricDataSameMetric-4        34.4          34.8          +1.16%

benchmark                                     old allocs     new allocs     delta
BenchmarkProcessMetricDataUniqueMetrics-4     0              0              +0.00%
BenchmarkProcessMetricDataSameMetric-4        0              0              +0.00%

benchmark                                     old bytes     new bytes     delta
BenchmarkProcessMetricDataUniqueMetrics-4     0             0             +0.00%
BenchmarkProcessMetricDataSameMetric-4        0             0             +0.00%