Handle S3 connection failure

[ramanathan106]

.flush() in _EventLoggerThread create a new connection each time, if there is fluctuation in connection boto3.client('s3', endpoint_url) 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.

[codecov-io]

Codecov Report

Merging #555 into master will decrease coverage by 0.2%.
The diff coverage is 41.66%.

@@            Coverage Diff             @@
##           master     #555      +/-   ##
==========================================
- Coverage   88.68%   88.47%   -0.21%     
==========================================
  Files          38       38              
  Lines        2774     2785      +11     
==========================================
+ Hits         2460     2464       +4     
- Misses        314      321       +7

Impacted Files	Coverage Δ
tensorboardX/event_file_writer.py	89.65% <41.66%> (-5.59%)	⬇️

---

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 d877dfe...cddad18. Read the comment docs.