Phases 5 & 7: FilterPhase - Refactor to Polars

[mattsan-dev]

Overview
The FilterPhase step now produces data using Polars LazyFrames, but the ConcatFieldPhase step still uses the legacy implementation and cannot process this new output. To keep the pipeline consistent and efficient, the FilterPhase phase must be updated to use the Polars version. This will keep existing behaviour while modernising the underlying process.

Tech Approach

• Update digital_land/phase/FilterPhase.py so it delegates to the Polars FilterPhase implementation.
• Implement the full FilterPhase logic inside digital_land/phase_polars/transform/FilterPhase.py.
• Convert existing FilterPhase rules into Polars lazy transformations.
• Ensure the FilterPhase phase accepts the LazyFrame produced by the new FilterPhase step.
• Keep the same public interface so no downstream code needs changing.

Acceptance Criteria/Tests

• FilterPhase phase accepts the FilterPhase phase LazyFrame without errors.
• Legacy FilterPhase behaviour is reproduced using Polars.
• Outputs match the legacy FilterPhase phase (apart from expected improvements).
• No Pandas or row‑based Python operations are used.
• Add test cases
• Refactored code complies with project formatting and style standards (black/flake8/pep)

Resourcing & Dependencies
Depends on ConcatFieldPhase, no external teams.

[mattsan-dev]

Refactored digital_land/phase_polars/transform/filter.py to use new Polars Lazy df method. Only rows where all applicable filters match are included e.g. prefix starts with "title" are kept, preserving existing business logic.

[lakshmi-kovvuri1]

@VENKAT-AVVARI-190825
We have added a unit test case for this one (tests/unit/phase_polars/transform/test_filter.py) please confirm that this is working and make any changes necessary.

Please also check that the refactored code complies with project formatting and style standards (black/flake8/pep) and make any necessary changes.