Make consistent the expectation around non-existent fragments directory

[jaraco]

Description

In the first commit, I update the documentation to reflect the current behavior, highlighting that a missing fragments directory is normal for a "create" operation and caused by the "build" operation.

In the second commit, I expand the documentation to include the case where "build" is invoked and the fragments directory is missing to be consistent with an empty directory or no directory at all. I also draft a likely way to achieve this behavior. Once tests pass, this approach closes #538.

Note that this change aligns the behavior with the currently prescribed expectation.

Checklist

• [x] Make sure changes are covered by existing or new tests.
• [x] For at least one Python version, make sure local test run is green.
• [x] Create a file in src/towncrier/newsfragments/. Describe your change and include important information. Your change will be included in the public release notes.
• [x] Make sure all GitHub Actions checks are green (they are automatically checking all of the above).
• [x] Ensure docs/tutorial.rst is still up-to-date.
• [x] If you add new CLI arguments (or change the meaning of existing ones), make sure docs/cli.rst reflects those changes.
• [x] If you add new configuration options (or change the meaning of existing ones), make sure docs/configuration.rst reflects those changes.

[adiroiban]

Thanks. Sorry for the extra trouble. I have merged it.

[iliakur]

@adiroiban you're gonna laugh, but I think we ran into this situation as well. I think we're gonna change our process for other reasons in a way that this isn't a problem. However, just in case, would you mind cutting a release with this in the next few days? Thanks a bunch!!

[adiroiban]

@iliakur here is the release PR https://github.com/twisted/towncrier/pull/566