Canonical Filament extraction engine

[kreneskyp]

Goal

Move Filament extraction policy currently owned by filament-parser-lib into quire-rs so Rust is the canonical engine for Python, WASM, IDE, and future sync consumers.

Scope

• Add a Filament-facing Rust module for one-document extraction.
• Support caller-provided ObjectType snapshots.
• Preserve Tier 1 and Tier 2 behavior from filament-parser-lib.
• Port frontmatter relationship sugar normalization, ix edge harvesting, edge dedupe, diagnostics, stable refs/IDs, and extraction error mapping.
• Return a JSON-serializable shape compatible with CoreExtractionResult.

Acceptance

• Rust tests cover representative Tier 1, Tier 2, unknown object, relationship sugar, ix links, duplicate edges, and extraction error cases.
• Output matches existing filament_parser.extract_core_data_from_markdown(...) on parity fixtures.
• No PGlite, Electron, HTTP, auth, or sync orchestration enters quire-rs.

Context

• Parent IDE context: agent-ix/filament-ide#24.
• This replaces the Python parser policy as the canonical extraction engine.