Tracking issue for shared source

[xxchan]

RFC risingwavelabs/rfcs#72

• feat(stream): add kafka backfill executor #14172
• feat: add kafka backfill frontend #15602
• schema change (Tracking: Alter Source/table with connector #11800, Tracking: alter source | (table with connector) #15582)
• scaling feat[DRAFT]: support scaling source backfill #16825
• CREATE TABLE (no change)
• pause/start from latest to avoid waste work
  • pause SourceExecutor before a Backfill is created. This means the SourceExecutor doesn't waste work, and the first MV (maybe we even want multiple MVs?) doesn't need backfill. feat: pause shared SourceExecutor until a downstream actor is created #16348
  • shared source: don't backfill for the first MV #16576
  • feat(source): let shared SourceExecutor start from latest instead of specified offset #16626
• Change the default behavior of create MV on source (backfilling) Use blocking (foreground) DDL for source and sinks #15587

testing

• more e2e tests improve script for running e2e source tests #15995
• longevity test
  • meta node OOM in longevity test with shared source #16577
• nexmark benchmark
  • https://buildkite.com/risingwave-test/nexmark-benchmark/builds/3711
• test: add some test for split change #15994
• a test idea: random gen data in kafka -> sink to kafka. Check result is the same. May also refer to mv backfill tests