feat: optional disable column stats

[parisni]

What is the purpose of the pull request

This PR makes sync faster and reduces memory footprint by adding a shared source-side option to skip column stats extraction:

xtable.source.skip_column_stats=true

The flag now works consistently for Hudi, Delta, and Iceberg sources.

In practical terms, this is intended for heavy tables where stats extraction is a bottleneck.
Example from large-sync behavior: a job that previously took around 6 hours and required >64 GB Xmx can be reduced to around 1 hour with about 10 GB Xmx.

Brief change log

• Added a shared source config (xtable.source.skip_column_stats) instead of per-format keys.
• Wired skip-column-stats behavior into all three source implementations:
  • Hudi source
  • Delta source
  • Iceberg source
• Kept required row-count behavior intact so downstream sync logic remains correct.
• Added handling for zero-row files where needed to avoid incorrect stats behavior.
• Improved naming consistency from “skip stats” to “skip column stats”.
• Added integration coverage for source format × sync mode × skip flag combinations.

Verify this pull request

This change added tests and can be verified as follows:

• Added parameterized integration test in ITConversionController for:
  • source format: Hudi / Delta / Iceberg
  • sync mode: Incremental / Full
  • xtable.source.skip_column_stats: true / false
• Verified with:
  • mvn -pl xtable-core -Dtest=ITConversionController#testSkipColumnStatsAcrossSources test
• Additional compile validation:
  • mvn -pl xtable-core -DskipTests compile

Trade-offs

• When xtable.source.skip_column_stats=true, column stats are not extracted or propagated.
• This reduces sync cost significantly, but column stats-dependent optimizations may be unavailable.
• As a result, query performance may be reduced for some workloads.