Refocus all skills on wall-clock build time as primary metric

[AvdLee]

Summary

• Establishes wall-clock build time (how long the developer waits) as the primary success metric across all skills, starting from AGENTS.md
• Adds blocking-vs-parallel heuristics to the orchestrator so it stops ranking parallelized compile hotspot fixes above serial bottlenecks
• Requires every recommendation to include a plain-language wait-time impact statement ("Expected to reduce your clean build by ~3s" or "Reduces parallel compile work but unlikely to reduce your build wait time")
• Adds disclaimers to timing-summary tables explaining that aggregated task totals can exceed wall-clock due to Xcode parallelism
• Updates verification/reporting in the fixer to lead with wall-clock deltas and honestly label cases where task metrics improved but wait time did not

Context

User feedback showed the skills were great at diagnosing issues (SwiftUI hotspots, build settings, SPM pinning) but over-promised impact by communicating in cumulative task time. This led to many source-level fixes that reduced compiler workload without actually reducing build wait time -- in one case adding 0.1s to the build despite large hotspot reductions.

Files changed (10)

• AGENTS.md -- wall-clock-first principle
• skills/xcode-build-orchestrator/SKILL.md -- prioritization rewrite, impact language, final report
• skills/xcode-build-orchestrator/references/orchestration-report-template.md -- timing disclaimers, wait-time impact examples
• scripts/generate_optimization_report.py -- timing-table disclaimers, wait_time_impact field rendering
• skills/xcode-compilation-analyzer/SKILL.md -- parallel workload labeling
• skills/xcode-project-analyzer/references/project-audit-checks.md -- wall-clock qualifier
• skills/spm-build-analysis/references/spm-analysis-checks.md -- wall-clock qualifier
• skills/xcode-build-fixer/SKILL.md -- wall-clock-first reporting
• references/benchmark-artifacts.md -- wall-clock vs cumulative distinction
• references/recommendation-format.md -- new wait_time_impact field

Test plan

• Run the orchestrator against a real project and verify the optimization plan includes wait-time impact statements per recommendation
• Verify timing-summary tables display the parallelism disclaimer
• Confirm the fixer reports results as "Your build now takes X.Xs (was Y.Ys)" format
• Check that compile hotspot recommendations are labeled as parallel workload when appropriate