Fix: add batch limit for stackdriver exporter

[mayurkale22]

Fixes #634

These changes have all been tested locally against a real stackdriver exporter and been confirmed to be working.

[mayurkale22]

/cc @mluggy Please review.

[mayurkale22]

@draffensperger Please review when you have a time.

[ofpiyush]

Can we merge this and get a release?

[philicious]

I'd like to mention another, related problem with exporting Metrics to Stackdriver:

Rate at which data can be written to a single time series: one point each 10 seconds

(used to be even only 1 per minute in December 2019 when I run into this problem)

its still ridiculously low for per-request metrics and also causes millions of API errors, eating up memory and CPU in your service if you dont implement some throttling yourself

this caused us to not even start using Stackdriver Monitoring for custom metrics but ended up using Prometheus again.

[6utt3rfly]

Is there a reason this isn't released yet? I'm running into the same issue as #634

[d-goldin]

Would be quite nice if this got some attention again. Kinda makes it barely usable with stackdriver.

[mamidon]

We're seeing this issue ourselves, is there any workaround available?

[philicious]

I guess there won't be much love for opencensus anymore as it's deprecated in favor of opentelemetry

[mamidon]

Bah, that would probably explain it. Of course the GCP docs I read must've been out of date. Thanks for the FYI.