core: griffe-based bounded context introspection

[absoludity]

Follows #106 . Main difference from #56 is using griffe rather than our own code for most of the AST parser. I've left comments/questions for things from #56 that are unclear to me.

Doctrine tests need to scan a codebase and discover its bounded contexts,
use cases, entities, and protocols — without importing any of the code being
scanned. Importing would require all dependencies to be installed and would
conflate scanning with execution.

This PR introduces static analysis via griffe,
a production-grade library used by mkdocstrings. It replaces what would
otherwise be a custom AST parser (the big PR #56 carries one at ~750 lines).
Griffe handles the full surface area of Python syntax edge cases we'd
eventually hit; we get that for free.

The `FilesystemBoundedContextRepository` is the primary output: given a
project root and search path, it discovers all bounded contexts and returns
the typed models from PR 1. The `FileSolutionConfigRepository` reads the
`[tool.julee]` section from `pyproject.toml` so the scanner knows where
to look.

Part of a stack

1. Domain models and constants
2. This PR — griffe-based bounded context scanner
3. Introspection use cases
4. Use case doctrine test

[absoludity]

Shouldn't this file be in .../repositories/**file/**introspection/init.py like the solution_config.py above? As in, the directory specifies that it's afile implementation, whereas currently it's sitting on its own?

It's been placed here because that's where it is in 56 but it looks (to me) like the whole of repositories/introspection should be under repositories/file/introspection

[absoludity]

Hrm - looks like introspection is explicitly not in file, even though it's a Filesystem repository?

[absoludity]

This is the main difference from the ast of #56 - using griffe to do most of the heavy lifting.