Better handle rows that break across splits, and other small related fixes

[srowen]

This attempt to address #398
See also #399

The change is I believe explained in comments below.

[srowen]

I'm not super happy about the hack here, but I could not find any other way around it. This is a potential correctness issue, so we need to do something. At least, I added more tests to exercise handling of rows that split across splits, and they still pass.

[HyukjinKwon]

Yea ... I don't like this hack too ... but seems no better way.

[HyukjinKwon]

Looks good if the styles and tests pass

[codecov-io]

Codecov Report

Merging #400 into master will decrease coverage by 0.05%.
The diff coverage is 83.33%.

@@            Coverage Diff             @@
##           master     #400      +/-   ##
==========================================
- Coverage   87.78%   87.73%   -0.06%     
==========================================
  Files          14       14              
  Lines         745      758      +13     
  Branches       64       65       +1     
==========================================
+ Hits          654      665      +11     
- Misses         91       93       +2

Impacted Files	Coverage Δ
...cala/com/databricks/spark/xml/XmlInputFormat.scala	92.53% <83.33%> (-0.86%)	⬇️

---

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 41d0d17...9584634. Read the comment docs.

[srowen]

Thanks @HyukjinKwon - we may want to cut an 0.6.0 release for this, plus the inferSchema change. I may need to learn from you how to do it.