(Re-)add `bowtie tui`, a TUI for interacting with implementations

We should write a nice TUI for executing Bowtie commands.

Some possible ideas are:

* an interactive evaluator of schemas/instances which keeps harnesses running in the background and streams over their outputs
* the ability to export/save schemas/instances/results as a report
* the ability to generate JSON Schema Test Suite test cases! And maybe even PRs
* ...

I'm labelling this as good-first-issue not because it's easy, but because it's self contained, and a fun project. Such a tui should use `rich`, and ideally be reasonably tested with automated tests. For anyone who picks this issue up, feel free to carve out only a minimal first version of this and to open further issues for any of the above further ideas (or your own).

(n.b. we used to have a similarly named `bowtie tui` which simply used [trogon](https://github.com/Textualize/trogon/), but which essentially was just a tool for assembling CLI invocations that as far as I know saw little actual use, had some bad bugs like #853 and certainly wasn't as useful as a true TUI.)

Start by reviewing the existing Bowtie commands and the former trogon-based `bowtie tui` behavior. Define a minimal first version rather than implementing every proposed idea, use `rich`, and add automated tests. Done means the TUI provides a useful initial way to interact with Bowtie implementations.

Written by the indexing model from the issue text.