Use incremental Redis scans for cache maintenance

[Apeksha9389]

Redis maintenance currently issues KEYS, which scans the whole database in one blocking command. Traverse incremental SCAN pages instead for sync/async cache clearing, calculation-flag clearing, and stale-entry removal. Preserve prefix filtering and existing warning behavior, and process delete/pipeline work per returned page.

Fixes #405; addresses item 10 in #367.

Validation:

• Redis test suite: 51 passed, 1 skipped (live Redis service unavailable; in-memory client tests ran).
• Six new regressions fail on master and pass here. They cover all six methods, an empty intermediate scan page, a nonempty final page, and prohibit KEYS calls.
• Additional integration checks with fakeredis and its async client: 350 entries per client, flag reset, stale deletion, full clear, and unrelated-key preservation passed. No new test dependency is required by this PR.
• Repository Ruff and formatting checks passed; git diff --check passed.
• Mypy has the same three pre-existing SQL backend errors as master with the installed dependencies; no new diagnostics.

Prepared with AI assistance; behavior and regressions were verified locally.