feat(codegen-kotlin)!: native arrays + text default by derivation (dbColumnType slim-and-derive Phase 1)

[dmealing]

Kotlin (Exposed) port of dbColumnType slim-and-derive — Phase 1. Derives what the metadata already says (field.subType + isArray + maxLength) instead of declaring it via @dbColumnType, aligning the Exposed output with the canonical TS Postgres DDL.

What changed (KotlinTypeMapper + Exposed table codegen)

text is the no-maxLength default

• field.string with no @maxLength now derives Exposed text(...) — matching the canonical TS text — instead of the old varchar(255).
• field.string with @maxLength: N → varchar(N) (the >4000 → text cutoff is preserved).
• stringMaxLength (which defaulted to 255) is replaced by stringMaxLengthOrNull (absent ⇒ null ⇒ text).

@kind:text hack removed

• The unregistered @kind attribute (KIND_TEXT / ATTR_KIND) that forced text(...) is gone. It was an unregistered attr that fails strict verify; with text now the no-maxLength default, the hack is unnecessary.

Native arrays are the DEFAULT (derived from isArray)

• field.string + isArray:true → array<String>("col", TextColumnType()) (native text[]).
• field.uuid + isArray:true → array<java.util.UUID>("col", UUIDColumnType()) (native uuid[]) — previously field.uuid + isArray fell through to the scalar uuid() column.
• The uuid_array / text_array / text @dbColumnType branches are removed (they derived the same thing the default now produces).
• isArrayResolved resolves array-ness through the extends super chain, so an object.projection field keeps inheriting its base field's array-ness (the projection-inheritance contract the regression test guards).
• kotlinTypeName stays the scalar/element mapper; the List<…> wrapping is applied by the entity/payload generators per isArray (avoids a List<List<…>> double-wrap). The native-array column DDL is emitted only in exposedColumnSpec.

Unchanged

• uuid (string → uuid column, String native) and jsonb (cross-port: verify+fix field.string @dbColumnType:jsonb → validator/DTO typing in C#/Python/Java/Kotlin (TS fixed in #97) #98 parsed-value open bag) escape hatches.
• timestamp / timestamp_with_tz (that's Phase 2 — native-type-changing).

Test evidence

• codegen-kotlin: 255 / 0 / 0. Updated KotlinTypeMapperTest (no-maxLength→text; string/uuid isArray→native array; @kind:text path removed; +maxLength→varchar), KotlinProjectionExtendsInheritanceTest (tags uses isArray:true instead of @dbColumnType:text_array), KotlinExposedTableGeneratorTest.
• integration-tests-kotlin: 92 / 0 / 0 — the Exposed write codec round-trips against Testcontainers Postgres.
• Snapshot single-entity-primitives/.../AuthorTable.kt regenerated: bio (no @maxLength) → text("bio"), proving the derived output.

Dependency on the Java agent

The shared CoreDBMetaDataProvider.java value-set is intentionally NOT modified here — removing the DB_COLUMN_TYPE_UUID_ARRAY / _TEXT_ARRAY / _TEXT constants from the cross-port provider is the Java agent's task. The Kotlin mapper no longer references those constants, so it compiles whether or not they've been removed yet. Once the Java agent removes them, an author can no longer write @dbColumnType: uuid_array/text_array/text (loader-rejected); until then those values are simply ignored by the Kotlin generator (arrays/text derive correctly regardless).

Part of dbColumnType slim-and-derive Phase 1

🤖 Generated with Claude Code

[dmealing]

Folded into the integrated Phase-1 PR (feat/dbcolumntype-slim-derive) which merges all five ports atomically — that's the one that lands. Keeping this open for per-port review history.

[dmealing]

Landed via the integrated Phase-1 PR #121 (squash-merged to main).