feat(scripts): add resumable bounded log scanner

[yzwooyi]

Summary

Issue #83 needs a client-side workaround for wide eth_getLogs requests. This adds a Python standard-library reference scanner that stores raw logs and its cursor in SQLite, commits each successful prefix atomically, and resumes safely after interruption.

• Freezes latest to a numeric end block and verifies chain ID and end-block hash on resume.
• Splits range-limited or timed-out requests adaptively without skipping blocks; bounded retries cover transient transport, HTTP 429/5xx, and Arc's concurrency limit.
• Validates the complete response batch, deduplicates by (blockHash, transactionHash, logIndex), and rolls back rows and cursor together on write failure.
• Documents the client-side workflow and adds dedicated Python CI.

This addresses #83 as a client workaround and does not change the RPC server. Related: #79 contains a simple pagination snippet; this adds an executable crash-resumable tool.

Validation

• python3 -m unittest discover -s scripts -p test_scan_logs.py -v — 25 tests passed under Python 3.12.14, including localhost HTTP transport fixtures.
• python3 -m py_compile scripts/scan_logs.py scripts/test_scan_logs.py — passed.
• git diff --check — passed.
• Live comparison: blocks 60786744..60786844 in 17-block chunks returned 1,836 raw logs exactly matching the direct RPC fixture (0 missing, 0 unexpected).
• A live from-genesis request returned provider error 4444 (pruned history unavailable), which is correctly treated as fatal rather than a range timeout.

No Rust or protocol changes are included. Full Rust and contract builds were not run because this is a client-script change; dedicated Python CI remains pending.