Preserve routine statement boundaries across SQL body forms

[minleejae]

PostgreSQL SQL RETURN functions and MySQL routines with a single SQL body could absorb the next statement into their raw declaration. Compound bodies could also stop at an inner CASE ... END or procedural terminator.

Extract routine boundary tracking from captureFunctionBody so simple bodies end at their own semicolon and nested blocks remain intact. Preserve the existing token-list API and quoted-body handling. Regression tests cover utility and direct parser calls, following-statement preservation, PostgreSQL SQL bodies, MySQL control flow/labels/handlers, and existing Oracle-style declarations.

Validation: complete Gradle check passed (8,992 tests, 0 failures), plus 87 original/toString/deparser executions on PostgreSQL 18.6, MySQL 9.7.2 and MySQL 26.7.0 with matching results and statement boundaries.

Performance smoke checks used the existing JMH SQL corpus and configured scripts with two forks and GC allocation profiling. Timing samples were noisy; existing-corpus allocation remained approximately 7.77 MB/op on the final refactor and at the base. An opaque MySQL body containing 5,000 nested IF blocks retained its following SELECT and completed without pathological growth.

Routine boundary transitions are separated into header, statement-boundary, block and control-flow helpers. The exact PMD NPathComplexity rule reported by Codacy (threshold 200) now passes without suppressions; the original violation was reproduced independently.

[manticore-projects]

Good one! Thank you very much for working on this!