feat(audio): list available TTS voices

[localai-org-maint-bot]

Description

Adds GET /v1/audio/voices and /audio/voices so clients can discover named voices for installed TTS models. The optional model query filters the response. The endpoint respects per-user model allowlists and resolves aliases.

Model configs can declare tts.voices entries with name, language, and gender metadata. Pocket TTS receives a built-in catalog, including pinned backend variants. The change also updates route discovery, API instructions, Swagger, authorization, configuration metadata, tests, and TTS documentation.

Closes #11604

Verification

• go test ./core/config -ginkgo.focus='TTSVoicesForModel'
• go test ./core/http/endpoints/localai -ginkgo.focus='TTSVoicesEndpoint'
• go test ./core/config/meta ./core/http/auth
• go test ./core/http/routes -run '^$'
• jq -e '.paths["/v1/audio/voices"].get.responses["200"]' swagger/swagger.json
• git diff --check

The full routes suite was not run because one existing spec requires a rootless Docker/Postgres provider that is unavailable in this worktree.

Notes for Reviewers

The endpoint returns only models with known voice metadata when unfiltered. A filtered installed model without metadata returns an empty voices list.

Signed commits

• Yes, I signed my commits.
• Documentation updated (docs/content/) for user-facing changes, or not applicable