Log SenderExceptions the first time they occur in a row

[pschichtel]

Which problem is this PR solving?

• Issue Log errors that occur during ThriftSender's flush #728

Short description of the changes

• From looking at the code, currently the SenderExceptions occurring in the Sender actually don't get logged at all. The RemoteReporter merely took the number of lost spans, but otherwise discards the exception. This change tracks if the previous command of its kind failed (using set of command classes that saw an error previously) and logs out 2 messages:

1. A message on the first failed execution
2. A message on the first successful message after it previously failed

Alternative approach

Alternatively I think we could check if the SenderException was caused by a SocketException and either:

• Log only the first SocketException
• Log only the first SocketException of its kind (comparing messages).

[codecov]

Codecov Report

Merging #729 into master will increase coverage by 0.26%.
The diff coverage is 100.00%.

@@             Coverage Diff              @@
##             master     #729      +/-   ##
============================================
+ Coverage     88.76%   89.02%   +0.26%     
- Complexity      596      598       +2     
============================================
  Files            73       73              
  Lines          2242     2251       +9     
  Branches        289      291       +2     
============================================
+ Hits           1990     2004      +14     
+ Misses          159      155       -4     
+ Partials         93       92       -1     

Impacted Files	Coverage Δ	Complexity Δ
...egertracing/internal/reporters/RemoteReporter.java	90.62% <100.00%> (+0.96%)	7.00 <0.00> (ø)
...gertracing/internal/reporters/LoggingReporter.java	90.90% <0.00%> (+9.09%)	5.00% <0.00%> (+1.00%)
...rtracing/internal/reporters/CompositeReporter.java	100.00% <0.00%> (+28.57%)	7.00% <0.00%> (+1.00%)

---

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 9b0ed16...a3c28e7. Read the comment docs.

[jpkrohling]

LGTM!