[NOTICKET] refactor(catalog): single SAMPLE_LIMIT constant shared across introspection paths

Hoist the per-column sample cap into catalog.introspect.base.SAMPLE_LIMIT so the
tabular and DB introspection paths reference one source of truth instead of two
separate literals (tabular head(3) + extractor SAMPLE_LIMIT=3) that can drift.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

src/catalog/introspect/base.py

@@ -9,6 +9,11 @@ from abc import ABC, abstractmethod
 
 from ..models import Source
 
+# Max sample values stored per column (down from 5 — token cost: sample values
+# are fed to the planner prompt). Single source of truth for every introspection
+# path (tabular files + DB), so the cap can never drift between them.
+SAMPLE_LIMIT = 3
+
 
 class BaseIntrospector(ABC):
     """Abstract base. Subclasses: DatabaseIntrospector, TabularIntrospector."""

src/catalog/introspect/tabular.py

@@ -26,7 +26,7 @@ from src.middlewares.logging import get_logger
 
 from ..models import Column, ColumnStats, DataType, Source, Table
 from ..pii_detector import PIIDetector
-from .base import BaseIntrospector
+from .base import SAMPLE_LIMIT, BaseIntrospector
 
 logger = get_logger("tabular_introspector")
 
@@ -198,7 +198,7 @@ class TabularIntrospector(BaseIntrospector):
             (document_id, sheet_name, col_name) if sheet_name else (document_id, col_name)
         )
 
-        sample_raw = series.dropna().head(3).tolist()
+        sample_raw = series.dropna().head(SAMPLE_LIMIT).tolist()
         sample_values: list[Any] | None = [_normalize(v) for v in sample_raw] or None
 
         is_numeric = pd.api.types.is_numeric_dtype(series)

src/pipeline/db_pipeline/extractor.py

@@ -12,12 +12,14 @@ import pandas as pd
 from sqlalchemy import Date, DateTime, Float, Integer, Numeric, inspect
 from sqlalchemy.engine import Engine
 
+from src.catalog.introspect.base import SAMPLE_LIMIT
 from src.middlewares.logging import get_logger
 
 logger = get_logger("db_extractor")
 
 TOP_VALUES_THRESHOLD = 0.05  # show top values if distinct_ratio <= 5%
-SAMPLE_LIMIT = 3             # sample N rows per column (down from 5 — token cost)
+# SAMPLE_LIMIT (sample N rows per column) is the shared cap defined in
+# catalog.introspect.base — one source of truth for both introspection paths.
 
 # Dialects with a single-statement CTE that survives `pd.read_sql`. On these we
 # fold the stats and sample queries into one round-trip per column. MySQL <8 and