feat(atomic-a11y): add merge-shards for parallel a11y report aggregation

[y-lakhdar]

TL;DR

Merge shard JSON reports from parallel CI runs into a single consolidated a11y report.

Context

When Storybook tests run with --shard=N/M, each shard writes its own a11y-report.shard-N.json. This PR adds the merge step that combines them.

CI (parallel shards)
  ├─ shard 1/3 → a11y-report.shard-1.json   ─┐
  ├─ shard 2/3 → a11y-report.shard-2.json   ─┤──→  mergeA11yShardReports()  →  a11y-report.json
  └─ shard 3/3 → a11y-report.shard-3.json   ─┘
                                                     • Deduplicate components across shards
                                                     • Sum violation/pass/incomplete counts
                                                     • Union criteria coverage + affected components
                                                     • Recompute summary from merged data

How the shard merge works

When Storybook tests run with --shard, each shard writes its own a11y-report.shard-N.json. The merge step recombines them into a single a11y-report.json.

mergeA11yShardReports() — entry point

1. Scan the output directory for files matching a11y-report.shard-{N}.json
2. Read & validate each shard file in parallel (Promise.allSettled — invalid shards are skipped with a warning)
3. Call mergeComponents() → mergeCriteria() → createSummary()
4. Take report metadata (product, versions, etc.) from the first shard, stamp a fresh reportDate
5. Write the merged a11y-report.json

mergeComponents() — deduplicate by component name

• First shard wins as the base record for each component
• Subsequent shards sum counts (storyCount, violations, passes, incomplete, inapplicable)
• Union criteriaCovered sets and incompleteDetails arrays
• Output sorted alphabetically by name

example
Shard 1 tested atomic-search-box with commerce stories, shard 2 tested it with search stories:

Shard 1:                                  Shard 2:
├─ name: "atomic-search-box"              ├─ name: "atomic-search-box"
├─ storyCount: 3                          ├─ storyCount: 2
└─ automated:                             └─ automated:
   ├─ violations: 1                          ├─ violations: 0
   ├─ passes: 12                             ├─ passes: 8
   ├─ incomplete: 0                          ├─ incomplete: 1
   ├─ inapplicable: 4                        ├─ inapplicable: 3
   ├─ criteriaCovered: ["1.4.3"]             ├─ criteriaCovered: ["1.4.3", "2.1.1"]
   └─ incompleteDetails: []                  └─ incompleteDetails: [{ruleId: "color-contrast", ...}]

Merged:
├─ name: "atomic-search-box"
├─ category: "search"                                       ← backfilled from shard 2 ("unknown" → "search")
├─ storyCount: 5                                            ← 3 + 2
└─ automated:
   ├─ violations: 1                                         ← 1 + 0
   ├─ passes: 20                                            ← 12 + 8
   ├─ incomplete: 1                                         ← 0 + 1
   ├─ inapplicable: 7                                       ← 4 + 3
   ├─ criteriaCovered: ["1.4.3", "2.1.1"]                   ← union of both sets
   └─ incompleteDetails: [{ruleId: "color-contrast", ...}]  ← concatenated

mergeCriteria() — deduplicate by criterion ID, two passes

1. Pass 1 — merge from shard criteria lists: union affectedComponents across shards for each criterion ID
2. Pass 2 — infer from merged components: if a component covers a criterion ID that no shard's criteria list included, synthesize a new criterion entry from WCAG metadata (logs a warning with the count of inferred criteria)

• Output sorted by numeric criterion ID (1.1.1 < 1.4.3 < 4.1.2)

example
Pass 1 — merge from shard criteria lists:

Shard 1 criteria:                         Shard 2 criteria:
├─ id: "1.4.3"                            ├─ id: "1.4.3"
│  └─ affectedComponents:                 │  └─ affectedComponents:
│     ["atomic-search-box"]               │     ["atomic-search-box", "atomic-result-list"]
└─ id: "2.1.1"                            └─ (no 2.1.1 entry)
   └─ affectedComponents:
      ["atomic-search-box"]
After pass 1:
├─ id: "1.4.3"
│  └─ affectedComponents: ["atomic-result-list", "atomic-search-box"]  ← union, sorted
└─ id: "2.1.1"
   └─ affectedComponents: ["atomic-search-box"]

Pass 2 — infer from merged components. Suppose atomic-result-list covers criterion "4.1.2" (from its criteriaCovered array), but no shard had a "4.1.2" criteria entry:

After pass 2:
├─ id: "1.4.3"  (from shards)
├─ id: "2.1.1"  (from shards)
└─ id: "4.1.2"  ← synthesized from WCAG metadata
   ├─ ...
   └─ affectedComponents: ["atomic-result-list"]
   [merge-shards] 1 criteria were inferred from component coverage.

PR Chain (4 of 7)

#	PR	Description
1	#7111	Package scaffolding
2	#7122	Shared types, constants, utilities
3	#7123	VitestA11yReporter + wiring
4	#7137	Shard merging ← this PR
5	#7124	OpenACR report generator
6	#7125	CLI scripts
7	#7117	Weekly a11y scan workflow

KIT-5469