Preserve LCOV source-path whitespace

[hzw0813]

Summary

• stop trimming LCOV SF: source-path records before repository-path resolution
• preserve leading/trailing ASCII and Unicode whitespace as literal path characters so coverage for one path cannot alias another changed file
• add regressions for trailing ASCII space, trailing NBSP, and leading ASCII space
• regenerate tracked dist and document the evidence-boundary fix

Why

parseLcov() used record.slice(3).trim(). That normalization changes LCOV path identity: for example, SF:src/value.js  (NBSP) was interpreted as src/value.js. If the artifact then reported hits for changed lines, ProofDiff could attach coverage evidence to a different repository path than the one named by the artifact.

Coverage artifacts are explicitly user-supplied and not independently attested, but ProofDiff still must not strengthen them by rewriting path identity. The parser now preserves the SF: payload exactly and lets normal repository containment checks decide whether it matches a changed file.

Verification

• full npm test passed after the change
• clean npm run clean && npm run build passed
• final branch diff contains only src/coverage.ts, tests/coverage.test.ts, CHANGELOG.md, and regenerated dist/coverage.*; temporary updater files were removed