fix(storage): prioritize large coverage compaction queues

[fewensa]

Summary

• rank coverage-index-v2 queued chains by queue size hint instead of a boolean queue check
• limit each compaction wake to the largest queued chains so hot coverage buckets get repeated processing instead of being diluted by every chain

Validation

• cargo fmt -p datalens-cli --check
• CARGO_TARGET_DIR=/code/helixbox/datalens/target CARGO_INCREMENTAL=0 cargo test -p datalens-cli storage_compaction_chain_batch -- --nocapture
• CARGO_TARGET_DIR=/code/helixbox/datalens/target CARGO_INCREMENTAL=0 cargo clippy -p datalens-cli --lib --tests -- -D warnings
• git diff --check