Fix CRAN testing issues with R-devel

[jayqi]

This PR resolves two issues we had with CRAN testing our 0.3.0 submission.

---

test_x64 for some Windows cases

• Added case for test_x64 to .BuildTestLib for some Windows cases

We had the following ERROR in our 0.3.0 submission to CRAN for Windows.

This is because we aren't catch the test_x64 case, which seems to be when Windows is using 64-bit R and explicitly calling it out in the directory names.

From R CMD check:

  Warning: invalid package 'd:/RCompile/CRANincoming/R-devel/pkgnet.Rcheck/tests_x64//baseballstats'
  Warning: invalid package 'd:/RCompile/CRANincoming/R-devel/pkgnet.Rcheck/tests_x64//sartre'
  Warning: invalid package 'd:/RCompile/CRANincoming/R-devel/pkgnet.Rcheck/tests_x64//milne'
  Warning: invalid package 'd:/RCompile/CRANincoming/R-devel/pkgnet.Rcheck/tests_x64'
  Error: ERROR: no packages specified

Updating .BuildTestLib to use R.home

Below is a reproduced error that we also got from CRAN testing on R-devel.

  * installing *source* package ‘milne’ ...
  ** R
  ** byte-compile and prepare package for lazy loading
  Error in rbind(info, getNamespaceInfo(env, "S3methods")) :
    number of columns of matrices must match (see arg 2)
  ERROR: lazy loading failed for package ‘milne’
  * removing ‘/tmp/RtmpMTUoXD/milne’
  Error in pkgnet:::.BuildTestLib(targetLibPath = Sys.getenv("PKGNET_TEST_LIB")) :
    Installation of packages in .BuildTestLib failed! (exit code = 1)

  WARNING: ignoring environment value of R_HOME ...  ...
  In addition: Warning message:
  In system(command = cmdstr, intern = TRUE) :
    running command '"/usr/bin/R" CMD INSTALL -l "/tmp/RtmpMTUoXD" /RPackage/pkgnet/pkgnet.Rcheck/00_pkg_src/pkgnet/inst/baseballstats /RPackage/pkgnet/pkgnet.Rcheck/00_pkg_src/pkgnet/inst/sartre /RPackage/pkgnet/pkgnet.Rcheck/00_pkg_src/pkgnet/inst/milne /RPackage/pkgnet/pkgnet.Rcheck/00_pkg_src/pkgnet' had status 1
  Execution halted

I'm not entirely sure what the problem was. I was able to reproduce this error just straight up R CMD INSTALL-ing milne on the R-devel docker image, though RD CMD INSTALL did not have any problem. This made me realize that .BuildTestLib is hard-coded to use R even for RD CMD CHECK.

Instead, we use file.path(R.home("bin"), "R") to look up the binary for the current R process. So if you call from an RD process, you will get the correct path to the r-devel binary.

https://stackoverflow.com/questions/33798115/command-to-see-r-path-that-rstudio-is-using

[codecov-io]

Codecov Report

Merging #154 into master will increase coverage by 0.1%.
The diff coverage is 100%.

@@            Coverage Diff            @@
##           master     #154     +/-   ##
=========================================
+ Coverage   84.55%   84.66%   +0.1%     
=========================================
  Files          10       10             
  Lines         913      913             
=========================================
+ Hits          772      773      +1     
+ Misses        141      140      -1

Impacted Files	Coverage Δ
R/testing_utils.R	80% <100%> (+2.85%)	⬆️

---

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 1a4a9fe...d3e5cfa. Read the comment docs.

[jameslamb]

LGTM

[bburns632]

Nice