Add instrumentation to the BatchSpansProcessor to count the number of dropped spans

[jkwatson]

related to #766

Is this the pattern that we think makes sense to use for instrumenting the SDK? Does this style of usage introduce a race condition in the creation of the SDKs? Does it mandate that the MeterSdk is created and installed before the TracerSdk is installed? Is there a better pattern to use?

[codecov-io]

Codecov Report

Merging #889 into master will increase coverage by 0.06%.
The diff coverage is 100%.

@@             Coverage Diff              @@
##             master     #889      +/-   ##
============================================
+ Coverage      82.3%   82.37%   +0.06%     
  Complexity      839      839              
============================================
  Files           111      111              
  Lines          2882     2893      +11     
  Branches        253      253              
============================================
+ Hits           2372     2383      +11     
  Misses          399      399              
  Partials        111      111

Impacted Files	Coverage Δ	Complexity Δ
...elemetry/sdk/trace/export/BatchSpansProcessor.java	90.55% <100%> (-0.83%)	7 <0> (ø)
...a/io/opentelemetry/sdk/metrics/LongCounterSdk.java	83.33% <0%> (+8.33%)	5% <0%> (ø)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b84bf39...a2bf9f5. Read the comment docs.

[bogdandrutu]

@jkwatson to answer some of your questions, the TracerSDK does not depend on the "BatchSpansProcessor" intentionally so there is no initialization race. The TracerSdk has minimal dependencies as well as the MeterSdk - exporters or configs are added later.