docs(getting-started): add Set Up a Model tutorial for API key and local paths

[Sayt-0]

Phase 4 (docs tutorial) of #3442, the last remaining phase. Closes #3442.

New getting-started page "Set Up a Model" walking both model-setup paths end to end, linked from the README, the quickstart, the site navigation, and the Phase 1 error messages.

Issue expectation mapping

Phase 4 expectation	Implemented by
New getting-started page "Set up a model (API key or local)" walking both paths end to end	docs/getting-started/set-up-a-model/index.md: comparison table, Path A (get an API key, store it through a secret source, verify with docker agent doctor, run), Path B (install Docker Model Runner, pull, verify, run), pointer to the docker agent setup wizard, doctor section
Linked from README	Install section paragraph and Documentation list
Linked from quickstart Options A/D	Note under Option A, sentence under Option D (the two spots issue scenario 8 called out)
Linked from the Phase 1 messages	Shared environment.ModelSetupDocsURL constant, appended to RequiredEnvError (model-credential variant), AutoModelFallbackError, and the first_available variant

Changes

File	Change
docs/getting-started/set-up-a-model/index.md	new page (canonical front matter, nav weight 25 between Installation and Quick Start)
docs/data/nav.yml	nav entry
README.md	links in install section and documentation list
docs/getting-started/quickstart/index.md	Option A note, Option D sentence
docs/community/troubleshooting/index.md	prose kept in sync with the new error-message links
pkg/environment/errors.go	ModelSetupDocsURL constant, link in RequiredEnvError model variant
pkg/config/auto.go	closing link in AutoModelFallbackError
pkg/config/first_available.go	closing link in the first_available missing-env error
pkg/{environment,config}/*_test.go	assertions that model errors carry the link and non-model errors do not

Message rendering after the change (combined with the Phase 3 wizard line):

No API key? Run a local model instead: docker agent run --model dmr/ai/qwen3 ...
(the model is pulled on first use; `docker model ls` shows models already pulled)
Or run `docker agent setup` to configure a provider or local model interactively.
Step-by-step model setup (API key or local): https://docs.docker.com/ai/docker-agent/getting-started/set-up-a-model/

Validation

Check	Result
go build ./...	ok
go test ./pkg/environment/ ./cmd/root/ and targeted pkg/config error tests	pass
golangci-lint run pkg/environment/ pkg/config/ cmd/root/	0 issues
markdownlint-cli2 (105 files)	0 errors
scripts/docs-check-canonical.sh (94 pages)	ok
lychee offline relative-link check	0 errors
Hugo site build (refLinksErrorLevel: ERROR)	ok, page emitted with nav entry

TestURLSource_Read_RejectsLocalAddresses and TestLoadExamples fail identically on a clean main in this environment (sandbox network restrictions, corrupted local DMR blob cache); unrelated to this change.

Issue status once merged

Phase	Status
1, actionable errors	shipped, #3452
2, docker agent doctor	shipped, #3489
3, guided setup wizard	shipped, #3499
4, docs tutorial	this PR