Support building/using PL/Java on Java 26

[jcflack]

This pull request enables PL/Java to build and run with Java 26, but does not bring any PL/Java API changes corresponding to any new features in the JDK 26 release notes. Some new features brought by Java 26's inclusion of the JDBC 4.5 maintenance release could merit future work:

• Verbatim syntax in JDBC SQL. JDBC now specifies that the recognition of JDBC escapes and the ? parameter placeholder should not take place between {\ and \} (spelled in a Java string literal as {\\ and \\}, respectively), and that \\ in such a region (\\\\ in a Java literal) represents a literal backslash.
  • This is a convenient way to write SQL that uses the ? character, which has become more common with recent additions to the language. However:
  • The specification is debilitatingly vague. It is silent as to whether those delimiters are scanned for blindly in the user-supplied SQL text, or with respect to existing SQL lexical rules for quoting and comments. Unless or until this oversight is remedied in an erratum or another maintenance release, grossly incompatible JDBC driver implementations are likely to result.
  • The way PGJDBC implements the feature could be looked to for an example, but PGJDBC, so far, does not seem to be implementing it, instead retaining a longstanding nonstandard convention where ?? represents a literal ?.
  • There is so much wrong with PL/Java's nativeSQL method that it would be difficult to add this feature without risk of changing behavior, be it even already-broken behavior, that is being relied on. A future major release that revamps the JDBC layer would be a better target for this change.
  • Although different from PGJDBC's ?? convention, PL/Java does provide a way to include ? in SQL when needed. It is a consequence of the failure of PL/Java's nativeSQL method to understand comments, while it does understand quoting. The string /*'*/?/*'*/ will therefore be seen by PL/Java as containing a ? that should be left alone because inside quotes, while PostgreSQL will recognize it as a ? surrounded by comments. This should be covered in PL/Java's documentation.
  • SQL in deployment descriptors is parsed by a different method that is differently broken: it does understand /* ... */ comments (and also // comments, even though those aren't a thing), which would seem to make the /*'*/?/*'*/ technique unusable. But because this method only recognizes comments that follow whitespace, the technique can still be used as long as no whitespace precedes it.
• New JDBC types DECFLOAT and JSON.
  • These new members of java.sql.Types and java.sql.JDBCType will be present when running on Java 26 or later, but PL/Java has no new behavior tied to those types.
• AutoCloseable implemented by several JDBC interfaces.
  • The functionality is supplied by default methods on those interfaces, requiring no new support in PL/Java. However:
  • The change was incompatible with PL/Java's old internal implementations of Blob and Clob, so those implementations are now different internally.
  • Those legacy implementations were only partial to begin with, and are probably little used, if at all.
  • Examples/regression tests have been added to capture what did and didn't previously work and ensure the new internal implementation is at least as complete as the old one.
  • A future major release that revamps JDBC support should replace these Blob and Clob implementations completely with versions that actually manipulate PostgreSQL large objects. That would bring the behavior in line with PGJDBC.

Naturally, no patch will begin using Java 26 features within PL/Java itself until a future major release adopts a newer minimum build version.