test(spp_change_request_v2): cover fail-closed/edge branches flagged by codecov on batch 2 (field_mapping no-selection; conflict_mixin dynamic-duplicate) [from #422]

[gonzalesedwin1123]

From the codecov patch-coverage audit of batch-2 PR #422 (comment: patch coverage 91.9%, 16 lines). The audit confirmed no security-critical enforcement line is untested — the 16 misses are defensive branches, latent-feature sub-cases, and mechanical async/demo lock reroutes. This issue tracks adding tests for the two Group-B lines worth pinning; the rest are defensive/mechanical and left uncovered by design.

Severity: Low (test coverage only)

Add tests

1. spp_change_request_v2/strategies/field_mapping.py:34 — return mappings.browse(), the fail-CLOSED path of _effective_mappings for a dynamic-approval type with no selected_field_name (applies nothing). The adjacent unmapped-selection branch (line 35) is tested; the no-selection branch is not. Add a test: a dynamic-approval field_mapping CR with an empty selected field applies/previews nothing.

2. spp_change_request_v2/models/conflict_mixin.py:528 — any_similar = True, the fuzzy-match sub-branch of dynamic duplicate detection (returns 80.0). Reached only by two dynamic-approval field_mapping CRs with the same changed-field set where a field differs but is "similar". Latent today (no shipped CR type enables dynamic approval) but worth a regression test alongside the existing dynamic-duplicate tests.

Explicitly NOT worth a test (documented for the record)

conflict_mixin.py:331/337/522 (missing-detail edge + non-existent-field skips), change_request.py:692 & change_request_detail_base.py:110 (recordset-normalization branch of _normalize_frozen_value), assign_program.py:65 (empty-program_id skip), and the eight spp_programs/spp_farmer_registry_demo _acquire/_release_operation_lock reroute call sites (async/demo paths the unit-test DB doesn't drive; the is_locked write guard itself is tested).

References

Codecov audit comment on #422; internal/plans/batch2-fix-review.md.