nargs=-1 works with envvar

[rch-liu]

This PR aims to fix two bugs:

1. Initially combining nargs=-1 and envvar would yield an empty tuple where the expected behaviour of nargs=-1 is to take on an unlimited number of arguments passed in envvar. @amy-lei and I added a fix so that now combining nargs=-1 with envvar should output all the expected values from envvar rather than an empty tuple.

2. We also added error catching for nargs > len(envvar) where we would raise: Got unexpected extra arguments and when nargs < len(envvar) we would raise: argument [argname] takes [nargs] values.

The changelog is also updated with our new changes. Let us know what you think @davidism!

fixes #729

[amy-lei]

We discovered that nargs was also not enforced when the values came from default, so we addressed/added tests for it as well.

[davidism]

Refactored the tests to use parametrize. Simplified the nargs validation so it uses the same message for any combination, rather than reusing the argument message, which didn't really fit.