Chi v0.1.0 - streaming corpus primitives

Chi's first public release provides four Zig 0.16.0 library modules for bounded, checkpointable corpus work: deterministic ingestion, streaming text cleanup, document filtering, and exact/approximate deduplication.

Validation: zig build test passes on Windows and Linux CI.

Scope: pre-1.0 library components, not a turnkey distributed pipeline. NFC/NFKC normalization, concrete language/safety classifiers, database drivers, distributed deduplication backends, and performance claims are explicitly out of scope.

Licensed under Apache-2.0. Developed under Cameron Emery's direction with material AI assistance.