also check for module wrappers in 'ModulesTool.exist' method

[boegel]

fix for #2601 (clearly still WIP, enhanced test will fail in current status)

[vanzod]

Tested with R-3.5.1-foss-2018b.eb. Java/1.8 is correctly recognized as installed.

[boegel]

Some tests are failing because of problems with the tests themselves, which I'll fix, but one failing test (test_stale_module_caches) highlights a problem that should be fixed: we're just checking for an existing .modulerc that defines a wrapper, but we're not checking whether the underlying module being wrapped actually exists.

Fixing that is fairly straightforward: check which module file is being wrapped, and then check whether that exists before spitting out a True for the existence check of the wrapper. This situation is probably mostly academic in the sense that it's unlikely to happen in practice (unless the module file that the wrapper refers to is removed, of course).

There's a complication though... The wrapped module is specified with the 'short' module name (e.g. Java/1.8.0_181) rather than the full module name (e.g. Core/Java/1.8.0_181), so we need to "translate" the short name to the full one based on the provided full name of the wrapper module (e.g. Core/Java/1.8).

Fixes for both issues coming up in next push (cfr. commits 06621bf + 525d92e)...

[boegel]

@vanzod This should be good to go now...