[SPARK-9905][ML][Doc] Adds LinearRegressionSummary user guide

[feynmanliang]

• Adds user guide for LinearRegressionSummary
• Fixes unresolved issues in [SPARK-9906] [ML] User guide for LogisticRegressionSummary #8197

CC @jkbradley @mengxr

[SparkQA]

Test build #41713 has finished for PR 8491 at commit 4e5aaeb.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):
  • public class LinearRegressionWithElasticNetExample

[mengxr]

LGTM. Merged into master and branch-1.5. Thanks!

Not related to this PR, the major issue with this page is missing documentation about input columns and output columns of linear models, and there are no example code to demo them.

Another question is whether we should keep the main class in Java example code. I think most of our Java example code do not have it, and assume jsc and sqlContext are provided. This could simplify the example code a little bit.

[feynmanliang]

@mengxr WRT 1, completely agree and created SPARK-10333

WRT 2, I agree we should be consistent throughout the code examples. However, I've felt awkward writing the Java examples at times because unlike Scala and Python there is no REPL which injects the context variables in and allows us to just copy/paste the example in to run. Keeping the main class makes it easy to paste the file, compile and JAR it, and use spark-submit to run the code example.