fix: Recommend --tag when publishing prerelease npm modules

[tmcw]

The --tag argument to npm publish allows you to publish a prerelease without the prerelease assuming
the 'latest' tag and causing users to download the prerelease by default. Instead, we recommend
--tag prerelease by default when --prerelease is set in the standard-version call, so to
install a prerelease version of the module, users will install module@foo or specify the exact
version of the prerelease version.

Fixes #183

[tmcw]

Quick notes:

• Considered a few approaches for what to name the tag, but prerelease seems to make sense. standard-version --prerelease without an arg generates, for instance, 1.0.0-0, so simply tagging the semver tag (npm install foo@0) doesn't work that well. Similarly, next would imply a major version bump, I think, so prerelease is the closest to the desired meaning.
• There isn't test coverage for output. Adding it would be a more invasive change, so I don't include that in this PR.

[coveralls]

Coverage increased (+0.0007%) to 99.614% when pulling fda4e04 on tmcw:advise-prerelease-tag into dfd1d12 on conventional-changelog:master.

[stevemao]

Thanks a lot! Could you add a simple test for that?

[tmcw]

Can do! I added the ability to test for output, by adding a Symbol key to the args object. It's the least invasive way and should avoid futurebugs around any new arguments, including arguments named messages. Open to other approaches if anyone sees other drawbacks to that testing approach.

[bcoe]

@tmcw thanks for this 👍 unfortunately (fortunately) I just finished a pretty big overhaul of standard-version so I think you might need to rebase.

Would love your thoughts on this old issue, regarding how we should generate a CHANGELOG for pre-releases.

[stevemao]

@tmcw Is there a way to use stdout just like other tests?

[tmcw]

Superseded by #196 which uses stdout to make testing much simpler.