test: cover inner PROCTRAN HWPT/XRTY catch in CRECUST and UPDCUST

[a2chang]

Closes #35.

What

Per-repository unit coverage for the inner PROCTRAN-insert catch in
CrecustRepository.createCustomer(...) and UpdcustRepository.updateCustomer(...).
That catch (added in #34) wraps non-retryable PROCTRAN failures as
CbsaAbendException("HWPT", "<PROGRAM> failed to write the audit trail.")
and re-throws SQLSTATE 40001 failures unchanged so CrdbRetry can retry
them. Until now both behaviours were exercised only indirectly via
DbcrfunRepositoryUnitTest (same catch pattern) and integration tests
(happy-path only).

How

The inner block sits inside dsl.transactionResult(configuration -> ...)
and uses DSL.using(configuration) to build the txDsl. Static-mocking
DSL.using is not available (subclass MockMaker doesn't support
mockStatic), so each test now hands the lambda a real jOOQ
Configuration backed by:

• org.jooq.tools.jdbc.MockConnection + a MockDataProvider that:
  • CRECUST: returns (CUSTOMER_COUNT=0, CUSTOMER_LAST=0) for the
    CONTROL SELECT ... FOR UPDATE, 1 row affected for the CUSTOMER
    insert and the CONTROL update, and throws the seeded
    SQLException on the PROCTRAN insert.
  • UPDCUST: returns one synthesised CUSTOMER row for the
    SELECT ... FOR UPDATE, 1 row affected for the CUSTOMER update,
    and throws on the PROCTRAN insert.
• A small inline ExecuteListener that calls Spring's
  SQLStateSQLExceptionTranslator.translate(...) and re-throws via
  ctx.exception(...). This mirrors what
  JooqAutoConfiguration#DefaultExceptionTranslatorExecuteListener
  installs in production (polish: jOOQ DataAccessException catches don't fire because Spring Boot translator substitutes Spring DAE #36 / fix(repos): catch Spring DataAccessException so CrdbRetry and PROCTRAN abends fire #37) — without it the test would surface
  a raw org.jooq.exception.DataAccessException, which the repository's
  catch (org.springframework.dao.DataAccessException) does not match.
  (Spring Boot's listener class is package-private; we cannot reuse it
  directly, so the test reproduces its one-line behaviour.)

dsl.transactionResult(callable) on the outer mock is then stubbed to
invoke the callable with that real configuration — the lambda's
DSL.using(configuration) returns a real DSLContext whose statements
flow through the MockDataProvider.

Tests

Four new tests, two per repository:

• nonSerializationProctranInsertFailureWrapsAsHwptAbend — non-40001
SQLException (e.g. 23505) → CbsaAbendException("HWPT", "<PROG> failed to write the audit trail.").
• serializationProctranInsertFailureSurfacesAsXrtyAfterRetryExhaustion —
  40001 SQLException → after CrdbRetry exhausts, the outer catch
  converts it to CbsaAbendException("XRTY", "<PROG> aborted after exhausting Cockroach serialization retries.").

Local ./mvnw -B verify: 202/202 (was 198/198 — +4).

Notes

• No production code touched.
• UpdcustRepositoryUnitTest's class-Javadoc previously documented why
  the inner catch was not unit-tested; that disclaimer is now obsolete
  and replaced with the actual coverage description, matching
  CrecustRepositoryUnitTest.
• The per-SQLException failure injection is the smallest scaffold that
  reaches the catch without a real database. Existing
  *ServiceIntegrationTest#proctranInsertFailureSurfacesHwptAbend tests
  (added in fix(repos): catch Spring DataAccessException so CrdbRetry and PROCTRAN abends fire #37) keep the integration-level assertion that the listener
  is actually wired in production.

augment review

[augment-app-staging]

Review completed. No suggestions at this time.

Comment augment review to trigger a new review at any time.