Point documentation at docs.networktocode.com and fix site defects

[jvanderaa]

Fixes #425

The Read the Docs Community project was retired, so every documentation link in the README and package metadata returns a 404. This repoints them, and the mkdocs site_url, at the Read the Docs for Business site under docs.networktocode.com.

A once-over of the site turned up several other defects, most of them cookiecutter leftovers that used the hyphenated project slug where the Python package name was required:

• Broken header logo. The theme pointed at docs/assets/networktocode_logo.svg while the file lived in docs/images/, so the logo 404'd. Moved the file (theme chrome belongs in assets/; images/ is for content images embedded in pages).
• "Edit this page" 404'd on every page. edit_uri was edit/main/circuit-maintenance-parser/docs; the default branch is develop and docs live at docs/.
• The Code Reference was silently empty. generate_code_reference_pages.py globbed a circuit-maintenance-parser directory that does not exist, so the API docs the README advertises generated zero pages. It now globs the real package and renders as a single page under Developer Guide, using the existing mkdocs-gen-files — no new dependency.
• Removed the unused site_dir, which wrote the built site into a Nautobot-app static/ path this library does not have.
• Fixed a v2.10 release-note link that used underscores in the repo name.

mkdocstrings heading_level moves 1 → 2 so each module becomes a heading below the page title, which restores the in-page table of contents.

Enabling the code reference surfaced seven griffe docstring warnings that fail the strict docs build; these are fixed at the source. poetry.lock is unchanged — this PR adds no dependencies.

Note: this does not by itself make the docs reachable

The Read the Docs for Business subproject still needs to be marked public by an admin. Until that happens, docs.networktocode.com/projects/circuit-maintenance-parser/ redirects to a login and the README docs badge reads "unknown".

Upstream follow-up

Four of these defects come from the python template in networktocode-llc/cookiecutter-ntc, which uses project_slug where project_python_name (or a plain path) is required. Any repo generated from it likely has the same empty Code Reference and dead documentation links. Worth fixing upstream separately.

[glennmatthews]

So this changes from a code-reference/ directory with one file per module into a single code-reference.md with all modules documented in a single file. Is that intentional/desirable?

[glennmatthews]

Do any of these fixes need to be additionally applied to the .cookiecutter.json file to make sure they don't get overwritten next time the package is rebaked?