Add RichCliRunner for testing Rich prompt/output capture

[mirs002]

Summary

• add rich_click.testing.RichCliRunner, a click.testing.CliRunner subclass
• rebind Rich's global cached console during CliRunner.isolation() so output goes to Click's isolated stdout/stderr streams
• restore the previous global Rich console after invocation to avoid leaking test state
• export RichCliRunner from rich_click
• add regression tests covering prompt capture behavior with a preconfigured global Rich console

Why

When Rich's global console has already been initialized before CliRunner.invoke(), prompt / console output may go to the pre-existing console stream instead of Click's test-captured streams. This makes asserting Rich prompt output difficult in tests.

This change provides a dedicated runner for Rich-based CLIs that keeps Rich aligned with Click's test isolation.

Closes #232.

[mirs002]

Note on the failing test (3.10, 8.1.*, 14.*) job

Just a heads-up that the one remaining CI failure is unrelated to this PR.

The job fails with an ImportError in conftest.py when importing Typer — before any of the new code in this PR is even loaded:

from typer.testing import CliRunner as TyperCliRunner
...
TypeError: 'type' object is not subscriptable

This is caused by Typer 0.24.x (released recently) dropping support for the Python 3.10 + Click 8.1 combination. The same job was passing on the last push to main on Jan 31 — before Typer 0.24 was released.

You can verify this by re-running the main branch CI — it will reproduce the same failure independently of this PR.

Happy to update the PR once this is resolved upstream or the matrix is adjusted! 🙂

[dwreeves]

Thanks! I'll review when I get a moment.