Scope 5 Extension: User-Declared String Sentinel Support

[DEVunderdog]

Parent

#94

Overview

Parallel feature to numeric sentinel support (#94/#170–175), extended to String/Utf8 columns with replace semantics: when a user declares ProfileConfig.string_sentinels for a column, only those values are treated as effective nulls (case-insensitive); the hardcoded defaults ("NA", "NAN", "NULL", "NONE", "?") are not applied for that column. Columns with no declaration are completely unaffected — hardcoded defaults remain the fallback. Empty/whitespace strings are always effective null regardless of any declaration.

Child issues

• feat: ProfileConfig.string_sentinels declaration + StructuralProfileResult propagation #177 feat: ProfileConfig.string_sentinels declaration + StructuralProfileResult propagation
• feat: extend _resolve_effective_nulls with user-declared string sentinel replace semantics #178 feat: extend _resolve_effective_nulls with user-declared string sentinel replace semantics
• feat: MissingnessProfiler string sentinel replace semantics #179 feat: MissingnessProfiler string sentinel replace semantics
• feat: profiling/orchestrator.py MAR detection — string sentinel replace semantics #180 feat: profiling/orchestrator.py MAR detection — string sentinel replace semantics
• feat: splitting/_profile_signals.py string sentinel replace semantics #181 feat: splitting/_profile_signals.py string sentinel replace semantics
• feat: FittedImputer string_sentinels field, transform wiring, and serialization round-trip #182 feat: FittedImputer string_sentinels field, transform wiring, and serialization round-trip
• feat: ImputationOrchestrator.fit() string sentinel threading + end-to-end integration test #183 feat: ImputationOrchestrator.fit() string sentinel threading + end-to-end integration test

Blocked by

None — can start immediately