model_specs: correct descriptions that contradict the code

[CryptVenture]

Split out of #372 as requested. This PR is description text only; the strict contract fixes, the VoxCPM1 defaults, the ASR/aligner/codec declarations and the MiniMax declarations are separate PRs.

Nothing here changes a declared type, bound, default or required flag, so no request that was accepted or rejected before changes state.

The changes

Spec	Change	Code it matches
dramabox	five descriptions dropped the claim that their default "comes from config.json"	dramabox/assets.cpp reads config.json for the architecture only; the values named are the engine's own
fish_audio.reference_text	says it is required whenever the request carries inline reference audio	src/models/fish_audio/session.cpp:255-258 throws "<role> with inline reference audio requires reference_text option"
outetts.reference_text	same, for voice cloning	src/community_models/outetts/session.cpp:695-698 and 722-728 throw "OuteTTS voice cloning requires --reference-text"

Both reference_text options stay required: false on purpose: they are required only when reference audio is present, and the schema's required is a plain bool with no conditional form, so marking them required would break plain TTS with a built-in voice. The description is the only place that distinction can currently live.

Validation

python3 tools/check_loader_catalog_sync.py
# ok: runtime loaders, model_specs, and model_manager_v2 are in sync

audiocpp_model_manager list --repository-root .        # exit 0
# parses every on-disk spec through the C++ schema loader

Scope

Three spec files, 7 lines. The generated WebUI bundle (webui/native/dist/index.html) is deliberately not included: catalog.ts inlines model_specs/*.json at frontend build time and the bundle is not byte-reproducible, so regenerating it in each PR of this split would produce conflicts between them. Happy to send one bundle-regeneration PR once the series has landed.

[CryptVenture]

Closing for now to stay inside the 3-concurrent-PR policy (see the discussion on #422). Nothing is wrong with the change and CI is green on it; I will reopen it as review slots free.