The test suite now runs the doctests in doc/tutorial.rst and executes the
shell session in doc/commandline.rst, comparing the actual output of the
jsonpointer command with the documented one. Both run as part of the
regular unittest suite, so CI covers them on every supported Python.

This uncovered four outdated outputs in the tutorial, where the documented
dict key order did not match what set_pointer returns; these are fixed.

The doc sources are added to the sdist so the tests also run from it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PewzpkEmq7YeZd56gmKnqr

jsonpointer.py stays a single-file module. Type information is shipped as
a PEP 561 stub-only package (jsonpointer-stubs) inside the same wheel and
sdist. A CI job runs mypy's stubtest to keep the stubs in sync with the
implementation.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PewzpkKQeJ6P2vfXcywcMy

The distribution ships the runtime module alongside the stubs, so
'Stubs Only' misdescribes it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PewzpkKQeJ6P2vfXcywcMy

Add type hints via a jsonpointer-stubs package

…son-68sghl

# Conflicts:
#	MANIFEST.in

A "$ " line right after a command's output started the next command
without checking the previous one, so its output was never verified.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PewzpkEmq7YeZd56gmKnqr

Check that documentation examples run and produce the expected output