chore(release): bump version to 0.4.3 + CHANGELOG

0.4.2 was already published to PyPI on 2026-06-03; the accumulated
batch ads/adgroups features and bugfixes (#566–#572) ship as 0.4.3.

Co-Authored-By: Claude Opus 4.8 (1M context) <[email protected]>

refactor: reuse _route_cpa_flag in update CPA routing (#491 C11/4a) (#…

…546)

The TextCampaign and UnifiedCampaign update paths each inlined a 19-line
CPA-flag routing closure (`_route_update` / `_u_route_update`) whose body
was byte-identical — including the `if value is None: return (None, None)`
guard — to the module-level `_route_cpa_flag` helper that the add path
already delegates to via thin wrapper closures (`_route` / `_u_route`).

Replace both inline closures with the same thin-wrapper pattern, binding
each block's captured Search/Network routing subtypes and delegating to
`_route_cpa_flag`. The closure names and every call site are unchanged;
only the def bodies move to the shared helper.

Payload-neutral: the wrapper passes the same captured subtype variables,
so `(search_value, network_value)` is identical for every input. Verified
byte-identical dry-run payloads (TEXT + UNIFIED), green parity-gate +
dry-run snapshots, full pytest 2177 passed, ruff clean. Net -8 lines.

Scope note (#544): Finding 2 (per-side fallback blocks) is NOT consolidated
— add uses unconditional defaults while update uses `... if value is not
None else None` ("don't touch the field"), which is payload-bearing, not
duplication. Finding 4b (builder-call kwargs) is NOT consolidated — the 3
differing kwargs (budget_type/include_default/is_update) are payload-bearing
and a dispatch table would hide exactly the copy-paste differences the
parity-gate guards. Only 4a is genuine, safe duplication.

Closes #544

Co-authored-by: axisrow <[email protected]>
Co-authored-by: Claude Opus 4.8 (1M context) <[email protected]>

chore(vendor): update tapi-yandex-direct to 2026.5.29

chore(audit): re-run wire-shape scanner + triage 2026-05-30 (#460) (#465

)

Re-ran scripts/audit_wire_shape.py --all after the captcha gate cleared.
Findings dropped 174 → 121; captcha-blocked endpoints 33 → 6.

- audit_wire_shape.py: recognise a WSDL `docs` base (#463) and record it as
  v5_wsdl_group_ok instead of deriving malformed `…?wsdl/get` URLs or
  flagging the 14-21 KB XML as thin/unreachable. Regression test added.
- Refresh docs/audits/wire_shape.json + PROJECT_WIRE_SHAPE_AUDIT_2026-05-30.md
  (replaces the 2026-05-29 snapshot).
- Add WIRE_SHAPE_TRIAGE_2026-05-30.md classifying every non-OK finding.

Triage result: no real code↔docs drift. All non-OK findings are parser
heuristic false-positives (code_field_missing/docs_field_missing/
required_field_missing — fields like Action/CampaignIDS/Currency are real in
both CLI and contracts), cosmetic (live4_marker_absent), known group-level
pages (agencyclients ×2), a documented 404 (CheckPayment — no public docs),
or transient captcha on 5 v4 methods that were verified live via browser at
dg-v4/ru/reference/<Method>.html and match their contracts.

Closes #460

Co-authored-by: axisrow <[email protected]>
Co-authored-by: Claude Opus 4.8 (1M context) <[email protected]>

Release 0.3.16

Release 0.3.15: drop --currency from v4finance methods to match docs 1:1

Release 0.3.14

Release 0.3.13

Add AdExtensions CalloutFieldNames flag

Release 0.3.11