noNA conflict of assertthat and terra

[falkmielke]

The assertthat::noNA check used in assertions might be masked by the terra::noNA function.
In those cases, the not-too-instructive error message is thrown:

> noNA(join_mask)
Error: unable to find an inherited method for function `noNA` for signature `x = "logical"`

This happened to me when using modules of the watina package in a non-RStudio context; I think the roxygen::@importFrom requirements prevent the error, for the majority of users. (I cannot tell whether those implicit imports run correctly in debugging contexts.)
However, combined use of terra and watina is quite common, we can not ensure IDE 100%, and the change is minor effort, so I would argue for an explicit call of assertthat::noNA.

The suggested change is nothing more than that: all noNA calls are replaced by assertthat::noNA.

Note: this is supposed to merge into dev_nextrelease, I hope I did it right with the PR.

[florisvdh]

It's not strictly needed when using debug mode in R since the package namespace already determines which noNA() applies (background: https://r-pkgs.org/dependencies-mindset-background.html#function-lookup-inside-a-package). But at the same time it's best practice, as recommended by https://r-pkgs.org/dependencies-in-practice.html#sec-dependencies-in-imports. Thanks!