fix: integration test failures for futuresearch-python (#5284)

## Summary

Fixes two failing integration tests for `futuresearch-python`:

- **`test_rank_validates_field_in_response_model`**: The error message
was changed in #5200 from `Field X not in response model Y` to
`field_name 'X' not found in response_schema properties: [...]`, but the
test regex wasn't updated. Updated the regex to match.

- **`test_dedupe_identify_strategy_no_selection`**: The dedupe
ClickHouse dual-write sidecar (#4554) hard-coded `selected` into
`DEDUPE_CH_SCHEMA`, so the IDENTIFY strategy was incorrectly producing a
`selected` column in the parquet output (filled with `False`). The
Supabase write path already handled this correctly (verified by
`test_dedupe_finalize.py::test_identify_strategy_no_selected_column`).
Replaced the constant schema with `_build_dedupe_ch_schema(strategy)`
that omits `selected` for IDENTIFY, threaded `config.strategy` through
`_init_ch_dedupe_context`, and removed the `else: data["selected"] =
False` branch from `_ch_write_dedupe_results`.

## Test plan

- [x] `pytest tests/services/test_dedupe_finalize.py
tests/services/test_dedupe_strategy.py` (60 passed)
- [x] `pyright src/engine/services/dedupe/{service,finalize}.py` (clean)
- [x] `ruff check` and `ruff format --check` (clean)
- [ ] `futuresearch-python` integration tests pass in CI

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

Sourced from commit 56714f7d50b6a73ec364ebc0a809b9f1d6cb00f9

Author: jackwildman

tests/integration/test_rank.py

@@ -102,7 +102,7 @@ class WrongModel(BaseModel):
 
     input_df = pd.DataFrame([{"item": "A"}, {"item": "B"}])
 
-    with pytest.raises(ValueError, match="not in response model"):
+    with pytest.raises(ValueError, match="not found in response_schema properties"):
         await rank(
             task="Rank items",
             input=input_df,