refactor: decompose mapping.go's three highest-complexity functions

[yvonnedevlinrh]

Summary

Phase 2b of CRAPload fragility reduction (#166). Decomposes the three highest-complexity functions in internal/quality/mapping.go — the assertion mapping engine — into 9 independently testable helpers.

Changes

Decomposition

Function	Before	After	Helpers extracted
matchContainerUnwrap	50	8	collectTrackedVars, traceForwardDataFlow, matchTrackedInExpr
isTransformationCall	26	5	isByteLikeParam, isPointerDestParam, resolveCallSignature, findParamIndex
matchAssertionToEffect	25	5	matchDirect, matchIndirectRoot
Combined	101	18	9 helpers

Test gap coverage

• +25 new unit tests using synthetic AST via parseAndTypeCheck
• Filled 3 isTransformationCall gaps: io.Reader parameter, empty interface destination, mixed parameter ordering
• matchAssertionToEffect now has direct unit tests (previously zero)
• matchContainerUnwrap helpers now have unit tests (previously integration only)

Test infrastructure cleanup

• Refactored parseAndTypeCheck to delegate to parseAndTypeCheckWithFset (DRY)
• Added comment documenting intentionally ignored type-check errors

Files changed (3 source + 5 spec + 2 learnings)

AGENTS.md                                          (+1)  — Recent Changes entry
internal/quality/mapping.go                        (+408/-168) — 9 helpers extracted
internal/quality/container_unwrap_internal_test.go (+653/-12) — 25 new tests, DRY fix

Verification

go build ./...                              # PASS
go test -race -count=1 -short ./...         # 16/16 packages PASS
golangci-lint run                           # 0 issues

# Complexity targets
gocyclo internal/quality/mapping.go | grep -E 'matchContainerUnwrap|isTransformationCall|matchAssertionToEffect'
# 8 matchContainerUnwrap, 5 matchAssertionToEffect, 5 isTransformationCall

# Mapping accuracy ratchet
go test -race -count=1 -run TestSC003_MappingAccuracy ./internal/quality/...
# 84.7% (floor: 84.0%) — PASS

Note on traceForwardDataFlow complexity

traceForwardDataFlow landed at complexity 32 — higher than the design's ~12 estimate. This is the forward data-flow tracing loop moved intact from matchContainerUnwrap. The parent function dropped from 50 to 8 because the concerns are now isolated, but the extracted core loop retains inherent complexity from nested ast.Inspect closures, multi-iteration convergence, and transformation call detection. Further decomposition of traceForwardDataFlow is a candidate for Phase 2c.

Review council

• Spec review: 3/5 APPROVE, 2/5 REQUEST CHANGES → fixes applied → all APPROVE
• Code review: 4/4 APPROVE (Architect 9/10, Testing, Guard, Adversary)

Spec artifacts

Full OpenSpec change at openspec/changes/crapload-decompose-pr2b/:

• proposal.md — motivation + constitution alignment
• design.md — 5 design decisions with alternatives rejected
• specs/decompose-mapping.md — 7 ADDED, 3 MODIFIED requirements with 24 scenarios
• tasks.md — 11 tasks in 5 groups, all checked off