Cache metric ID and forward the metric info to series writer

[antekresic]

Description

This is an optimization to remove the need to fetch metric ID and
table name when storing new series data since we already fetch
this info in a prior step of the process.

Merge requirements

Please take into account the following non-code changes that you may need to make with your PR:

• CHANGELOG entry for user-facing changes
• Updated the relevant documentation

[JamesGuthrie]

There's no CHANGELOG entry, this change might be worth mentioning if it has an impact.

[antekresic]

@JamesGuthrie I've added an entry to CHANGELOG and update the benchmarks. I've left the old function in and added a new benchmark so we can compare the improvement.

This is what I get on my machine:

$ go test -bench=BenchmarkGetSeriesIDForKeyValueArrayNewSeriesNewLabelsBatch -run=XXX -log-level=error -benchtime=10000x
...
BenchmarkGetSeriesIDForKeyValueArrayNewSeriesNewLabelsBatch-12
Labels DB time 722.503306ms per item: 72.214µs N-labels 10005
DB time 4.892545455s per item: 489.254µs N 10000
   10000            563624 ns/op
...
BenchmarkGetSeriesIDForKeyValueArrayNewSeriesNewLabelsBatchOptimized-12
Labels DB time 729.144623ms per item: 72.878µs N-labels 10005
DB time 2.778612362s per item: 277.861µs N 10000
   10000            352828 ns/op
PASS

I can always remove the old if we want to do it, I'll let Mat take a look and advise there.

[cevian]

This is a very nice improvement. I think it's good to run the benchmark before and after the change and record diff in PR (as you have above). But I don't think we should keep the old functions around just for that purpose. Also I think that in a bunch of places (maybe all places) where you now pass around the metricID and tableName we should pass around a model.MetricInfo instead. That will make the code more future proof.