Connection Error Fix

[Saurav-D]

Adding on to the PR: #555.
Context: .flush() in _EventLoggerThread create a new connection each time, if there is fluctuation in connection S3 or GCS throws an error and since it is not handled the thread will hang and since the queue is full the training will also hang. The try block added will prevent the thread from getting stuck, instead it waits for the connection to appear again. Since it's a while loop the training wont resume till the connection is established again. Connection variable will make sure the print happens only once.

Refer this issue for more details: #606

I'm unsure if this is the right place to catch the error maybe it can be done individually in GCS and S3 writer.

[codecov-io]

Codecov Report

Merging #607 into master will decrease coverage by 0.17%.
The diff coverage is 56.25%.

@@            Coverage Diff             @@
##           master     #607      +/-   ##
==========================================
- Coverage   80.73%   80.56%   -0.18%     
==========================================
  Files          39       39              
  Lines        2824     2835      +11     
==========================================
+ Hits         2280     2284       +4     
- Misses        544      551       +7     

Impacted Files	Coverage Δ
tensorboardX/event_file_writer.py	89.74% <56.25%> (-5.54%)	⬇️

---

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 34d1616...7cab42f. Read the comment docs.