trying cleaner travis build with workaround for testthat bug

[goldingn]

again, opening PR to test new set up on Travis

[goldingn]

sorry, pr to wrong repo

[AugustT]

I saw that the .yml file for spocc was very different to ours but I'm afraid I know little about travis. It also looks like the ModuleHelp tests I just wrote are failing, but they are fine on my machine, BALLZ

[goldingn]

Yup, I prefer the metacran package builder, but couldn't get it to work because of this devtools bug. There's a fix suggested there so trying that here (or rather here).

If this works it will run faster and test on R-devel as well as the previous version of R. We could probably also set up CRAN and non-CRAN checks (and we can skip awkward tests in the CRAN versions),

That's a big if, last time I tried for many hours and then gave up after polluting the commit history

[goldingn]

well whatdya know, it (almost) worked!