v0.71.0

• Fix unbounded memory growth when stitching multi-line log events
  • Log entries spanning many lines, such as large auto_explain JSON plans, or a
    log_line_prefix mismatch that makes every line look like a continuation,
    previously caused excessive memory allocation and could OOM the collector
  • The continuation content stitched onto a single log entry is now capped at
    10MB, with any excess dropped and a warning logged
• Track temporary file usage per database
  • Collects temp_files and temp_bytes from pg_stat_database
• Track IO worker backends for Postgres 18+
• Report collector start time to the pganalyze server
• Fix collector hang on config reload
  • Both the connect and reconnect logic could get stuck waiting on Go channels
    that were not ready, causing reload (SIGHUP) to hang
• Fix schema collection ignore_schema_regexp filtering for large partitioned catalogs
  • Previously the column_stats query did not exclude catalog objects matched
    by the regexp
  • Also significantly speeds up relation_stats collection on databases with
    many partitions
• Prevent file and goroutine leaks during self-hosted log tailing
• Helm chart: Allow passing deployment annotations
• Update aws-sdk-go from v1 to v2
• Fix RPM installation on FIPS-enabled systems by using sha256 payload digests
• Update Go version to 1.26