feat: categorize Amazon refund items

[eshaffer321]

What changed

• Read Amazon's Return Center history and per-return status pages directly with the selected account's saved cookies.
• Capture authoritative order ID, RMA, return-created date, refund total, refund-issued date, status, ASIN, item name, and item price.
• Add itemize amazon returns -account <name> for read-only JSON inspection without initializing Monarch, SQLite, or an LLM.
• Integrate eligible Amazon refunds into normal sync and report categorized versus untouched refunds separately.
• Add a unique transaction matcher that rejects tied candidates.
• Normalize browser-exported quoted cookie values and Monarch category whitespace.

Safety behavior

• Only positive, non-pending, unsplit credits that are uncategorized or still in [TEMP] Amazon are eligible.
• A refund requires Amazon's explicit refund total, explicit issued date, RMA, and exactly one returned item.
• Multi-item returns without authoritative item-level allocation remain untouched.
• Same-date/same-amount credits are updated as a group only when Amazon record count equals Monarch credit count and every returned item independently resolves to the same category.
• Group notes explicitly state that the bank feed cannot identify the individual credit-to-item mapping.

Root cause

Amazon's normal order-payment transaction endpoint did not expose usable refund entries. The authoritative return/item mapping lives in Return Center, while the refund-issued date lives on each return status page. The issued-date sentence spans multiple HTML elements, and Monarch stored the temporary category with repeated whitespace, so both sources needed structured normalization.

Verification

• Mandatory failing regression tests were written before each fix.
• Pre-commit formatting, golangci-lint, go vet ./..., and go test ./... pass.
• go test ./... -race passes.
• go build -o itemize ./cmd/itemize passes.
• Full authorized account dry-run: Amazon refunds: Categorized=4 Left untouched=0.
• Production pass categorized the four target credits.
• A fresh Monarch read-back confirmed the final categories and truthful item/group notes.

Related: #72 improves Amazon setup cookie persistence. This branch is independently rebased directly onto origin/main v0.2.2 as requested.

[codecov]

Codecov Report

❌ Patch coverage is 80.57692% with 101 lines in your changes missing coverage. Please review.
✅ Project coverage is 64.29%. Comparing base (ba770cd) to head (d1aae63).

Files with missing lines	Patch %	Lines
internal/adapters/providers/amazon/returns.go	79.71%	26 Missing and 17 partials ⚠️
...ternal/application/sync/handlers/amazon_refunds.go	87.73%	10 Missing and 10 partials ⚠️
internal/application/sync/amazon_returns.go	79.56%	13 Missing and 6 partials ⚠️
internal/domain/matcher/matcher.go	73.33%	6 Missing and 6 partials ⚠️
internal/application/sync/orchestrator.go	0.00%	7 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #74      +/-   ##
==========================================
+ Coverage   62.10%   64.29%   +2.19%     
==========================================
  Files          46       49       +3     
  Lines        6040     6560     +520     
==========================================
+ Hits         3751     4218     +467     
- Misses       2035     2036       +1     
- Partials      254      306      +52     

Flag	Coverage Δ
unittests	64.29% <80.57%> (+2.19%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
internal/application/sync/types.go	31.87% <ø> (ø)
internal/application/sync/orchestrator.go	33.33% <0.00%> (-2.13%)	⬇️
internal/domain/matcher/matcher.go	81.39% <73.33%> (-8.85%)	⬇️
internal/application/sync/amazon_returns.go	79.56% <79.56%> (ø)
...ternal/application/sync/handlers/amazon_refunds.go	87.73% <87.73%> (ø)
internal/adapters/providers/amazon/returns.go	79.71% <79.71%> (ø)

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.