feat(evidence): boruna evidence inspect shows step output content

[escapeboy]

Summary

• Extends boruna evidence inspect to read outputs/<step_id>/result.json files for plaintext (non-encrypted) bundles
• Text mode: prints === Step Outputs === section with [<step_id>] <json> entries, truncated to 500 chars with ... (truncated) suffix
• --json mode: adds "step_outputs" key to the JSON object (map of step_id → parsed serde_json::Value; invalid JSON falls back to {"raw": "..."})
• Encrypted bundles without --decrypt now hint: "pass --decrypt --bundle-encryption-key <hex> to view step outputs" instead of just printing the algorithm info

Test plan

• inspect_shows_step_outputs_for_plaintext_bundle — creates minimal bundle dir with outputs/step_1/result.json, asserts step_id and value appear in stdout
• inspect_json_includes_step_outputs_for_plaintext_bundle — same bundle with --json, asserts step_outputs.step_1.value == 42
• All existing boruna-cli tests continue to pass (including inspect_does_not_leak_plaintext_without_decrypt_flag and inspect_decrypt_reveals_step_outputs_with_correct_kek)

🤖 Generated with Claude Code

[github-actions]

Bench compare

Threshold for regression: ≥ 10% slower mean.

Benchmark	Mean change	99% CI
compile_crud_admin_template	-1.59%	[-1.73%, -1.42%]
compile_medium_program	-2.17%	[-2.38%, -1.93%]
compile_small_program	-7.24%	[-7.63%, -6.74%]
evidence_build_5_steps	-2.13%	[-2.61%, -1.61%]
evidence_build_empty	-0.08%	[-0.46%, +0.28%]
evidence_verify_10_steps	+0.50%	[-1.24%, +2.29%]
evidence_verify_5_steps	-0.84%	[-3.99%, +1.71%]
vm_call_dispatch_loop/iters=1000	-3.16%	[-3.37%, -2.97%]
vm_call_dispatch_loop/iters=10000	-2.39%	[-2.48%, -2.30%]
vm_pure_loop/iters=1000	-2.61%	[-2.78%, -2.46%]
vm_pure_loop/iters=10000	-2.01%	[-2.10%, -1.91%]
vm_pure_loop/iters=100000	-2.50%	[-2.63%, -2.30%]
vm_record_loop/iters=1000	-0.68%	[-0.83%, -0.52%]
vm_record_loop/iters=10000	-1.70%	[-1.91%, -1.55%]

✅ No benchmark regressed past threshold.