feat(schema): accept PostgreSQL spellings of supported column types

[omer-cengel]

Summary

Schemas written for PostgreSQL often use the type names PostgreSQL itself documents, such as TIMESTAMPTZ, INT8, SERIAL8, or CHARACTER VARYING(20). sqlcj previously rejected these with Unsupported SQL column type. They now map to the same Java types as the spellings sqlcj already accepted, so an existing PostgreSQL schema no longer has to be rewritten for code generation.

Changes

• INT4 and SERIAL4 map to Integer; INT8 and SERIAL8 map to Long; INT2, SMALLSERIAL, and SERIAL2 map to Short.
• CHARACTER VARYING, CHAR, and CHARACTER map to String. PostgreSQL blank-pads CHAR(n) values, so ab stored in CHAR(3) reads back as ab .
• TIMESTAMP WITHOUT TIME ZONE maps to java.time.LocalDateTime and TIMESTAMPTZ to java.time.OffsetDateTime.
• Every serial spelling (SMALLSERIAL, SERIAL, BIGSERIAL, SERIAL2, SERIAL4, SERIAL8) is modeled as non-null. The integer aliases take their nullability from NOT NULL only.
• Matching stays case-insensitive and ignores type arguments, as before, and the rejection message for unsupported types is unchanged.
• Schema-parser tests cover each spelling, and a PostgreSQL integration test round trips a table declared entirely through these spellings via generated code.
• The PostgreSQL documentation lists the accepted spellings and no longer lists them as rejected.

Scope and non-goals

• Grammar-only synonyms (CHAR VARYING, DEC, NCHAR, BPCHAR), quoted or schema-qualified type names, the distinct "char" type, and array types remain unsupported.
• No new Java types, analyzer, generator, or runtime changes.