Add new print_traceback configuration option

[kdeldycke]

Allow the full traceback to be printed when Sphinx encounters an internal fatal error.

Purpose

• Improve in-situ debuguability of Sphinx.
• Improve bug reporting of third-party extensions.

Detail

For instance, when Sphinx is run in CI/CD (like GitHub actions workflows), we do not have access to the file-system. So when Sphinx encounter a fatal internal error, it just print the last frace of the traceback and dump the full traceback in a local temporary file.

This can produce error logs as uninformative as:

(...)
Exception occurred:
  File "<docs>", line 3, in <module>
AssertionError
The full traceback has been saved in /tmp/sphinx-err-wjogkybw.log, if you want to report the issue to the developers.
Please also report this if it was a user error, so that a better error message can be provided next time.
A bug report can be filed in the tracker at <https://github.com/sphinx-doc/sphinx/issues>. Thanks!

So by always printing the full traceback, this PR will improve identification of root causes, especially if they're coming from third-party plugins and extensions. I expect this change to have the nice side effect of improving bug reports to the community.

[AA-Turner]

Thank you for the very nice PR @kdeldycke, and sorry it has taken a month to review.

The only real question I have is on design -- is a config option the right choice here? Should we use an environment variable or command-line flag instead?

In other words, what are the scenarios for disabling this feature, and when are people most likely to do so? I am half-minded to unconditionally enable it, but there might be complaints of overly-long console output.

A