fix(storage): bound snapshot cleanup records

[fewensa]

Summary

• cap coverage v2 snapshot cleanup record generation per bucket
• stop listing snapshot heads once the cleanup batch cap is reached
• keep query-side latest snapshot lookup unchanged so read correctness and max-delta safeguards remain intact

Why

Production compaction can spend a long tick publishing many cleanup records from one old bucket. That delays hot bucket cleanup and leaves foreground queries hitting over-budget delta buckets. This change makes snapshot cleanup incremental so maintenance keeps rotating through buckets.

Verification

• cargo fmt --check
• cargo test -p datalens-storage --test maintenance test_compaction_coverage_index_v2 -- --nocapture
• cargo test -p datalens-storage test_s3_background_operations_have_shorter_timeout -- --nocapture
• cargo clippy -p datalens-storage --all-targets -- -D warnings

Safety

• does not raise or disable max_delta_objects
• does not skip coverage checks
• does not delete objects without existing cleanup records
• only batches cleanup publication work per tick