Port transactional storage access to Crawlee for Python: a request handler's storage writes should be buffered and committed atomically, so a handler that fails or gets retried leaves no partial writes behind. It should be enabled by default in every crawler and implemented at the storage level, so direct Dataset / KeyValueStore / RequestQueue calls are covered too, not just the crawling context helpers - which also absorbs the adaptive crawler's bespoke buffering.

Follow the JS implementation:

• Extend transactional storage access for general usage and enable it by default crawlee#3796
• feat!: Transactional storage access crawlee#3953

This will be a breaking change.

✍️ Drafted by Claude Code