Set _R_CHECK_FORCE_SUGGESTS_ to FALSE

[jeroen]

CRAN sets _R_CHECK_FORCE_SUGGESTS_=0 in their checks because some suggests might not be available on Windows, but the check still passes. I have the same problem with e.g. protolite: https://ci.appveyor.com/project/jeroenooms/protolite

* checking package dependencies ... ERROR
Package suggested but not available: 'RProtoBuf

I think it would make sense to default to _R_CHECK_FORCE_SUGGESTS_=0 on AV as well?

[krlmlr]

Interesting. But what are the effects if _R_CHECK_FORCE_SUGGESTS_=0 is the default and the tests are not prepared for missing suggested packages? The error message is rather self-explanatory now, and it's easy to fix on the client side.

[jeroen]

If this is the case you get a NOTE (rather than ERROR) during CMD check:

* checking package dependencies ... NOTE
Package suggested but not available for checking: ‘RProtoBuf’

And then if the package is actually required it just errors where the package is used:

Error in library("RProtoBuf") : there is no package called ‘RProtoBuf’

So I think it's rather obvious what the problem is.

[krlmlr]

I'd rather not set this globally, but add a pointer to this setting in the README and suggest using this in a build matrix.

[krlmlr]

OTOH, this is now the default setting for travis::use_tic(), so I guess it should be the default. @jimhester: Do/should we have _R_CHECK_FORCE_SUGGESTS_ off by default on Travis?

[krlmlr]

Fixed by documentation.