[FLINK-40453][postgres] Not load table schema when initializing record emitter

[naivedogger]

What is the purpose of this pull request?

Fixes FLINK-40453.

PostgresPipelineRecordEmitter currently builds all CreateTableEvents eagerly in its constructor (and again in the applySplit() override), which opens a JDBC connection and fetches every captured table's schema up front — before any split is processed. This adds startup overhead that scales with the number of captured tables. A table's schema is only needed when its first record / low-watermark / schema-change event is emitted, so loading everything at initialization is wasteful.

Brief change log

• Remove the eager generateCreateTableEvent() call from the constructor and drop the applySplit() override that pre-populates the cache.
• Resolve and cache each CreateTableEvent lazily on demand, falling back to a JDBC lookup only when the split state does not already carry the table schema.
• Behavior for snapshot / incremental / schema-change paths is unchanged.

---

Verifying this change

This change added tests and can be verified as follows:

• Added unit tests in PostgresPipelineRecordEmitterTest, asserting that constructing the emitter opens no JDBC connection and that assigning a snapshot split does not trigger loading of all table schemas.

Documentation

• Does this pull request introduce a new feature? (no)
• If yes, how is the feature documented? (not applicable)

---