test.data.table manual says to run in clean environment

[jangorecki]

closes #3705

[codecov]

Codecov Report

Merging #4352 into master will not change coverage by %.
The diff coverage is n/a.

@@           Coverage Diff           @@
##           master    #4352   +/-   ##
=======================================
  Coverage   99.61%   99.61%           
=======================================
  Files          72       72           
  Lines       13916    13916           
=======================================
  Hits        13862    13862           
  Misses         54       54           

---

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 b1b1832...244f317. Read the comment docs.

[mattdowle]

What does "clean environment" mean precisely? Should users rm(*), or does it mean "fresh new R session"? If the latter, what if the very reason the user wants to run test.data.table() is if they want to test in that particular session; maybe they think some packages might be causing bad interaction effects with data.table, or that their R session may be corrupted or something.

I really think #3705 should be fixed directly. I thought it already ran in its own independent scope, but clearly it must still be able to reach .GlobalEnv. If there's not a way to isolate it from .GlobalEnv, then test 1894 could be made robust to this.