fix(storage): shorten background s3 timeouts

[fewensa]

Summary

• keep foreground S3 operations at the existing 30s timeout
• shorten background S3 operations to 5s so compaction/cleanup cannot sit on slow RustFS calls for minutes
• add a unit test locking the foreground/background timeout split

Safety

• query/read correctness is unchanged: foreground store keeps the 30s timeout
• coverage max-delta guard is unchanged
• background maintenance is retryable and will continue on later ticks after transient timeout failures

Tests

• cargo fmt --check
• cargo test -p datalens-storage test_s3_runtime -- --nocapture
• cargo test -p datalens-storage test_s3_background_operations_have_shorter_timeout -- --nocapture
• cargo test -p datalens-storage --test maintenance test_compaction_coverage_index_v2 -- --nocapture
• cargo clippy -p datalens-storage --all-targets -- -D warnings