Search result models

[sigma67]

Part of #307. Base branch: ytmusicapi-2. Read the contributor guide from #982 before starting.

The hardest Phase 2 issue and the one that depends on all three core entity models. Best picked up once #983–#985 have landed and a few other Phase 2 issues have shaken out the patterns.

Scope

• search in mixins/search.py — results across all filters: songs, videos, albums, artists, playlists, podcasts, episodes, profiles, community playlists, featured playlists, uploads
• get_search_suggestions, including detailed_runs=True, which currently returns list[str] | JsonList

Watch out for

• The return type is a union discriminated by resultType. Model it so consumers can narrow on resultType and get the right fields — a single model with every field optional would defeat the point of this whole effort.
• scope="library" and scope="uploads" return different shapes from the default scope.
• ignore_spelling and the top-result shelf produce entries that differ from the regular result rows.
• Podcast results without a browseId (test: skip podcast search results without a browseId #976) must stay tolerated.

Definition of done

• Each resultType has a model and narrowing works in mypy.
• get_search_suggestions is properly typed for both detailed_runs values, ideally via overloads like get_lyrics uses.
• Tests cover every filter.
• mypy --strict clean.