Use Spark DF/RDD APIs for sampling without replacement instead of SQL

[nhanitvn]

Description and related tickets, documents

• Also provide an optimized way of sampling for integer sample sizes using Spark DF sample and limit.
• The following error is fixed

Total size of serialized results of 793 tasks (2.0 GB) is bigger than spark.driver.maxResultSize (2.0 GB)

• This is to update the fix in https://github.com/ddf-project/DDF/pull/349 to the main dev branch

Reviewers: @hai-adatao @phvu @Huandao0812

Breaking changes & backward compatible issues

No

How to test

PR Progress

Make sure all checkboxes below are checked before merged

• [x] Branch is in format prefix/description (see this)
• [ ] Merge check has no conflicts. PR checks passed.
• [ ] Code review is done.

[nhanitvn]

retest this please

[nhanitvn]

retest this please

[hai-adatao]

Lots of PE and PyClient / RClient tests failed after this merge: https://ci.arimo.com/job/BE-CI-PE-test/1012/ https://ci.arimo.com/job/BE-CI-RClient-test/232/ https://ci.arimo.com/job/BE-CI-PyClient-test/290/ I'm gonna revert this