fix: silence find_header warn on optional Greeks columns (#472)

[userFRM]

Summary

Closes #472.

crates/thetadatadx/src/decode.rs::find_header emitted a
tracing::warn!("expected column header not found in DataTable", header=…)
on every optional column missing from a wire response. The Greeks family
splits the column set across endpoints — _first_order ships seven Greeks,
_second_order ships five, _third_order ships four — but the shared
GreeksTick schema carries the full 23-Greek union. The reporter saw
eight warn lines per row on a single option_snapshot_greeks_first_order
call from Python. Demote the diagnostic to tracing::trace! so genuine
schema-drift investigations are still reachable via
RUST_LOG=thetadatadx=trace while routine subset calls stay clean.

Required-column drift continues to surface as a typed
Error::MissingRequiredHeader from the generated parser, so the
silenced path is strictly the documented optional-column case.

Pin the per-endpoint vendor column lists for the four Greeks families
in the GreeksTick schema doc-comment, against the upstream OpenAPI
capture in scripts/upstream_openapi.yaml. The struct layout itself
is unchanged — every Greeks endpoint still returns Vec<GreeksTick> —
so the SSOT pickup is doc-only and there is zero generated-SDK drift.

Add three regression tests in decode.rs::tests driving
parse_greeks_ticks against _first_order, _second_order, and
_third_order wire shapes. Each asserts bit-exact decoded values for
the wire-present columns and 0.0 defaults for the documented gaps,
so a future regression of find_header back to tracing::warn! —
or any column-list drift in either direction — surfaces as a
behavioural test failure rather than as live log spam.

Per-endpoint vendor schemas pinned to:

• _first_order: items_option_snapshot_greeks_first_order — delta, theta, vega, rho, epsilon, lambda + IV pair
• _second_order: items_option_snapshot_greeks_second_order — gamma, vanna, charm, vomma, veta + IV pair
• _third_order: items_option_snapshot_greeks_third_order — speed, zomma, color, ultima + IV pair (vera not in third-order)
• _all: items_option_snapshot_greeks_all — full 23-Greek union
• _implied_volatility: items_option_snapshot_greeks_implied_volatility — handled by IvTick

Workspace 8.0.25 → 8.0.26, tdbe 0.12.6 → 0.12.7. CHANGELOG +
docs-site/docs/changelog.md kept byte-identical.

Follow-up: per-endpoint typed structs

The bug-report comment thread also asked for per-endpoint typed
return structs (GreeksFirstOrderTick, GreeksSecondOrderTick,
GreeksThirdOrderTick) so the type system reflects the wire schema.
Surveying the surface, that change ripples through ~30 codegen
tables (build_support helpers, FFI macros with hand-tuned padding,
hand-written C++ headers, Go FFI mirrors, three SDK regenerations,
the EndpointOutput enum + every consumer in CLI / MCP / server /
ffi). It deserves its own issue + PR rather than being bolted onto a
warn-silence fix; the schema doc-comment in this PR is the
prerequisite that nails down the column lists per endpoint, so the
follow-up can map each subset to a typed struct without re-deriving
the wire shape from scratch.

Test plan

• cargo fmt --all -- --check clean
• cargo clippy --workspace --all-targets -- -D warnings clean
• cargo test --workspace 0 failures across all crates
• cargo deny check clean (advisories + bans + licenses + sources)
• cargo run -p thetadatadx --bin generate_sdk_surfaces --features config-file -- --check no drift
• Three new regression tests in decode::tests exercising the
  three Greeks subset wire shapes
• Reporter on Issue fix(decode): silence find_header warning for optional columns #472 confirms log spam is gone after pulling
  v8.0.26