WIP: bd-164 timestamp-only export deduplication (~80% complete)

Implemented content hash-based deduplication to skip exports when only
timestamps changed. Core logic complete, needs export_hashes table wiring.

Completed:
- Added computeIssueContentHash() excluding timestamps
- Created shouldSkipExport() logic
- Updated export loop to skip timestamp-only changes
- Added hash.go with content hashing
- Extended Storage interface

Remaining:
- Complete export_hashes table migration
- Add SetExportHash/GetExportHash to interface
- Revert content_hash from dirty_issues approach
- Wire up hash persistence in export
- Testing

See bd-164 notes for details.

Amp-Thread-ID: https://ampcode.com/threads/T-d70657d1-4433-4f7e-b10a-3fccf8bf17fb
Co-authored-by: Amp <amp@ampcode.com>

internal/storage/sqlite/schema.go

@@ -120,6 +120,14 @@ CREATE TABLE IF NOT EXISTS dirty_issues (
 
 CREATE INDEX IF NOT EXISTS idx_dirty_issues_marked_at ON dirty_issues(marked_at);
 
+-- Tracks content hash of last export for each issue (for timestamp-only dedup, bd-164)
+CREATE TABLE IF NOT EXISTS export_hashes (
+    issue_id TEXT PRIMARY KEY,
+    content_hash TEXT NOT NULL,
+    exported_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    FOREIGN KEY (issue_id) REFERENCES issues(id) ON DELETE CASCADE
+);
+
 -- Issue counters table (for atomic ID generation)
 CREATE TABLE IF NOT EXISTS issue_counters (
     prefix TEXT PRIMARY KEY,