BenchmarkFit — fit a portfolio surface to a benchmark

BenchmarkFit (the comparison facet — CANONICAL_INTELLIGENCE_OBJECTS.md §9): fit a subject portfolio's weight vector against a benchmark surface at a common teo (the subject's latest teo ≤ as_of; the benchmark then at its latest teo ≤ the subject's teo — never peeking ahead). Returns active share, an active-weight RMS (a coarse tracking-error proxy — factor-based TE is a follow-on), overlap, and the top over/underweights. subject = a BW-* portfolio id (BW-FUND-…, BW-FILER-…, BW-ETF-…, BW-BENCH-…) or an ETF ticker (→ BW-ETF-{TICKER}); benchmark = a bw_bench_id or an alias (SPY, 70/30, …) — the static benches, free on the gateway plane — or a custom bench (billed $0.005/call, capability bench-active-custom, parity with per-entity fund reads): ff_own (free-float-cap weight of the subject's OWN holdings — conviction vs what market cap alone implies; held symbols without a valid cap are dropped and counted in benchmark_provenance, never synthesized), cell_<9-box slug> (e.g. cell_large-growth — the style cell's MV weight surface at the latest teo ≤ the subject's teo), or all (fan-out: SPY + ff_own + the subject's declared style cell when resolvable, one call, response { fits: [...], omitted: [...] }). Readiness gate: benchmarks under development (hollow trailing teos, shallow history, unverifiable stores) are blocked with 409 before billing; on all they are listed in omitted[] with reason under_development.