address all R CMD check notes

[aappling-usgs]

Leaving these unaddressed for what I think are good reasons:

• Won't fix require('rstan') call because requireNamespace isn't enough. rstan::stan_model() and/or rstan::sampling() can't find the rstan function cpp_object_initializer() unless the namespace is loaded. requireNamespace is somehow not doing this. An online discussion (without solution) is at https://stat.ethz.ch/pipermail/r-devel/2014-September/069803.html

* checking dependencies in R code ... NOTE
'library' or 'require' call to 'rstan' in package code.
  Please use :: or requireNamespace() instead.
  See section 'Suggested packages' in the 'Writing R Extensions' manual.

• Won't fix ::: calls to 'devtools:::github_resolve_ref' and 'devtools:::parse_git_repo' because (1) I'd have to copy devtools code to get the functionality I want without the ::: - I want to get the name of the most recent tag on a GitHub branch - and (2) these calls occur in an unimportant part of the package (.onAttach), are wrapped in a tryCatch, making it unlikely that changes to the unexported functions would damage anybody's workflow.

* checking dependencies in R code ... NOTE
Unexported objects imported by ':::' calls:
  'devtools:::github_resolve_ref' 'devtools:::parse_git_repo'
  See the note in ?`:::` about the use of this operator.

• Won't fix call to attach() in revise() because I really want the clean interface I get from using attach(), have tried and failed to achieve it using eval and evalq and with, and believe I am following the best practices outlined in ?attach.

* checking R code for possible problems ... NOTE
Found the following calls to attach():
File 'streamMetabolizer/R/revise.R':
  attach(specs, warn.conflicts = FALSE, name = "revisespecstemp")
See section 'Good practice' in '?attach'.