Add demo bisect seed history

[rameziophobia]

Summary

• add a documented commit history for exercising compare-bisect against the demo storefront
• introduce homepage visual-regression and performance-regression checkpoints as separate commits
• document clean checkpoints, route coverage, and fixture conventions

Stack

This PR is stacked on #160 so its review shows only the eight demo seed-history commits. Once #160 merges, this PR can be rebased or retargeted to main.

Review note

The granular commit sequence is part of the feature: compare-bisect needs distinct historical checkpoints to locate the seeded regressions. The final net diff is two files.

[claude]

Review

Verified the actual net diff (2 files, +42/-1) matches the PR description. Both changes look intentional and internally consistent with the stated bisect-fixture goal, but one maintainability concern stands out:

runMerchandisingWarmup reads as production code, not a fixture (demo-ecommerce/app/javascript/components/pages/HomePage.tsx)

The codebase already has a convention for this exact situation — ExperimentA11yRegressions.tsx names itself unambiguously and carries a doc comment: "Intentional experiment-only accessibility regressions used to exercise the compare report's 'new in experiment' UI." It's also visually marked (data-cy="experiment-a11y-regressions", dashed red border).

The new runMerchandisingWarmup() function has none of that: the name implies real business logic ("merchandising warmup"), there's no comment flagging it as a synthetic perf-regression fixture for git-bisect, and it produces no visible/DOM marker. A future contributor skimming HomePage.tsx (without having read docs/git-bisect-seed-history.md) could easily mistake this 450ms main-thread busy-loop for real logic worth "fixing" (which would defeat the bisect fixture) or worth optimizing/removing without realizing it's load-bearing for a specific historical checkpoint.

Suggest renaming to something self-evidently synthetic (e.g. simulateHomepagePerfRegression) and adding a one-line comment pointing at docs/git-bisect-seed-history.md, matching the ExperimentA11yRegressions precedent.

Everything else — the hero gradient color swap (visual-regression fixture) and the new docs file — is low-risk and consistent with the described bisect-seed-history design.

[claude]

This is a deliberate 450ms main-thread busy-loop planted as a git-bisect perf-regression fixture (per docs/git-bisect-seed-history.md), but nothing here signals that. The name reads as real merchandising logic, and there's no comment or marker (unlike ExperimentA11yRegressions.tsx, which is clearly named/commented/tagged as an intentional fixture).

Recommend renaming to something self-evidently synthetic (e.g. simulateHomepagePerfRegression) and adding a comment pointing at the bisect-seed-history doc, so a future reader doesn't mistake this for a real bug or delete it without realizing it's load-bearing for a specific checkpoint.