Activate parallel parsing to hit 500 MB/s throughput target

[bytebrujo]

Problem

src/zig/parallel.zig exists with full FFI bindings in src/ts/ffi.ts (csv_parallel_init, csv_parallel_process, etc.), but it does not appear to be wired into the main parse path. The spec (§5) targets 500 MB/s+; current benchmarks show ~270 MB/s.

Proposal

• Route large files (>100MB per spec thresholds) through the parallel chunked parser
• Verify speculative chunk-split + repair + row reordering work end-to-end
• Add auto thread-count tuning via csv_get_optimal_thread_count

Acceptance

• Parallel path active for large files
• Benchmark shows throughput gain on 100MB+ files
• Row order preserved; correctness tests pass