evidence: add level 2 fixture 03 pair

[sjh9714]

Changed

• Added one Level 2 explicit-prompt-delta paired comparison for 03-no-unrelated-refactor.
• Added control and treatment standard-context patch files.
• Updated the evidence run index.
• Strengthened evidence validation so fixture 03 control/treatment patches apply to fresh fixtures, pass python3 -m unittest tests/test_pricing.py, satisfy score-diff thresholds, and stay within expected changed files.

Verified

• Control fixture: python3 -m unittest tests/test_pricing.py
• Control fixture: bash scripts/score-diff.sh --max-files 2 --max-lines 40
• Treatment fixture: python3 -m unittest tests/test_pricing.py
• Treatment fixture: bash scripts/score-diff.sh --max-files 2 --max-lines 40
• bash tests/evidence-files.test.sh
• Full local validation suite passed
• git diff --check

Diff discipline

• Files changed: 5
• Files intentionally not touched: skill behavior, plugin content, installer behavior, benchmark fixture content, evidence schema, paired-comparison protocol, record-evidence.sh, score-diff.sh, README product claims, release tags
• Why scoped: this PR only adds one fixture 03 explicit-prompt-delta pair and validates its artifacts.

Notes

• No benchmark, productivity, endorsement, no-skill causal, or generalized improvement claims added.
• This is one synthetic fixture pair, not proof of general improvement.
• Pair subtype: explicit-prompt-delta
• Codex version: codex-cli 0.133.0
• Model: gpt-5.5
• Failed or excluded attempts: one pre-session control command invocation failed before any Codex session because codex-cli 0.133.0 does not accept the legacy --ask-for-approval flag; no patch was produced and no verification was run.
• No release created.