[SPARK-9834][MLLIB] implement weighted least squares via normal equation

[mengxr]

The goal of this PR is to have a weighted least squares implementation that takes the normal equation approach, and hence to be able to provide R-like summary statistics and support IRLS (used by GLMs). The tests match R's lm and glmnet.

There are couple TODOs that can be addressed in future PRs:

• consolidate summary statistics aggregators
• move dspr to BLAS
• etc

It would be nice to have this merged first because it blocks couple other features.

@dbtsai

[SparkQA]

Test build #41977 has finished for PR 8588 at commit 34107aa.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):
  • case class Instance(w: Double, a: Vector, b: Double)

[SparkQA]

Test build #41979 has finished for PR 8588 at commit c75ff92.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):
  • case class Instance(w: Double, a: Vector, b: Double)

[SparkQA]

Test build #41994 has finished for PR 8588 at commit 1614f22.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):
  • case class Instance(w: Double, a: Vector, b: Double)

[dbtsai]

Do we need standardizeLabel? I think without regularization, with/without standardization will not change the solution.

[mengxr]

We don't need it but I think it is useful to list the values explicitly here.

[feynmanliang]

LGTM, did not check low level implementation

[feynmanliang]

jenkins test this please

[SparkQA]

Test build #42145 has finished for PR 8588 at commit c2ec746.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds the following public classes (experimental):
  • case class Instance(w: Double, a: Vector, b: Double)

[mengxr]

Merged into master. I will make follow-up PRs to do the refactoring.