[SPARK-19158][SPARKR][EXAMPLES] Fix ml.R example fails due to lack of e1071 package.

[yanboliang]

What changes were proposed in this pull request?

ml.R example depends on e1071 package, if it's not available in users' environment, it will fail. I think the example should not depends on third-party packages, so I update it to remove the dependency.

How was this patch tested?

Manual test.

[srowen]

Yeah, why did this ever depend on e1071? doesn't seem essential.

[SparkQA]

Test build #71204 has finished for PR 16548 at commit 5e2dd6b.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[yanboliang]

Merged into master and branch-2.1. Thanks for reviewing.

[holdenk]

So there is something similar in the fulltests for R ./R/pkg/tests/fulltests/test_mllib.R (found while working on packaging).

[yanboliang]

@holdenk Could you let me know where we meet similar issue in the fulltests? AFAIK, we test functions in e1071 only when it was installed on that node, like following:

if (requireNamespace("e1071", quietly = TRUE)) {
    expect_error(m <- e1071::naiveBayes(Survived ~ ., data = t1), NA)
    expect_equal(as.character(predict(m, t1[1, ])), "Yes")
  }