refactor: lazy sync requests and batched payment store updates

[ben-kaufman]

This PR:

1. Splits wallet_sync_request into separate full-scan and incremental methods so only the needed request type is built per sync cycle
2. Introduces a WalletSyncRequest enum to carry the chosen sync variant through the pipeline
3. Spawns primary and additional wallet syncs concurrently in a single JoinSet instead of running primary sequentially first
4. Batches update_payment_store_for_all_transactions and write_node_metrics into a single call after all wallet updates are applied
5. Fixes a pre-existing bug where additional wallet sync timestamps were set even when apply_update_for_address_type failed

QA Notes

Testing

• cargo test passes
• cargo clippy clean
• Verify Electrum sync path works with multiple address types
• Verify Esplora sync path works with multiple address types
• Confirm payment store is updated correctly after sync

Notes

• No API changes — internal refactor only, no binding regeneration needed
• The apply_update and apply_update_for_address_type methods now require callers to call update_payment_store_for_all_transactions afterward (documented via doc comments)

Made with Cursor