Add `--filter-featured` flag to filter lyrics sections by artist attribution

[shimpeiws]

Problem

Genius lyrics contain section headers like [Verse 2: Drake] with artist attribution, but processor.py strips these without parsing. Featured artist vocals are counted as primary artist vocabulary.

Additionally, remove_section_headers=True in GeniusClient.__init__() causes lyricsgenius to strip headers before BarScan can parse them.

Proposed Changes

src/barscan/genius/client.py

• Change remove_section_headers=True to False (line 64) - BarScan's clean_lyrics() already handles removal

New file: src/barscan/analyzer/sections.py

• SectionHeader model (frozen): section_type, artist_name: str | None, raw
• SECTION_HEADER_WITH_ARTIST_PATTERN = re.compile(r"\[([^\]:]+?)(?::\s*(.+?))?\]")
• parse_section_header(), extract_section_headers()
• is_artist_match() - case-insensitive substring matching, unknown artists default to inclusion
• filter_lyrics_by_artist() - keeps primary artist sections and unlabeled sections

src/barscan/analyzer/models.py

• Add filter_featured_sections: bool = Field(default=False) to AnalysisConfig

src/barscan/analyzer/processor.py

• Add primary_artist and featured_artists params to preprocess()
• Call filter_lyrics_by_artist() before clean_lyrics() when config enables it

src/barscan/analyzer/frequency.py

• Add primary_artist and featured_artists params to analyze_text(), pass through to preprocess()

src/barscan/cli.py

• Add --filter-featured option (default off)
• Pass primary_artist=artist_data.artist.name to analyze_text()

Tests

• New tests/test_analyzer/test_sections.py: header parsing, artist matching, lyrics filtering
• tests/test_analyzer/test_processor.py: preprocess with filtering
• tests/test_cli/test_commands.py: --filter-featured flag

Considerations

• Section header format varies (community-edited on Genius)
• Japanese headers supported (e.g., [1番: KOHH])
• Best-effort, not guaranteed

Dependencies

No dependency on Issue 1 (independent feature).