convert c++ exceptions to r errors in module .External entry points

[kevinushey]

Fixes #1495.

CppMethod__invoke, CppMethod__invoke_void, CppMethod__invoke_notvoid, and class__newInstance in src/module.cpp are .External entry points that can throw C++ exceptions -- most notably CHECK_DUMMY_OBJ's Rcpp::not_initialized, the designed error for using a module object created via the dummy-object path (new(Class) with no arguments, for a class without a default constructor). With no BEGIN_RCPP/END_RCPP in these functions, the exception escapes into R's C evaluator uncaught and std::terminate() aborts the whole R session, instead of raising the intended R error. See #1495 for details.

Changes:

• src/module.cpp: wrap the four entry points in BEGIN_RCPP/END_RCPP, so exceptions are converted to R errors as elsewhere in Rcpp. On the not_initialized path this also unwinds the local XPtr objects properly before the R error is raised.
• inst/tinytest/test_module.R: regression test -- calling a method on an uninitialized ModuleRandomizer must raise the "C++ object not initialized" R error. Without the fix this test aborts the tinytest process.

One pre-existing quirk surfaced while testing (left unchanged, noting it here): when a dummy object is garbage collected, the reference class finalizer calls class_<T>::run_finalizer(), whose XP(object) rejects the dummy pointer (an environment), printing Error: Expecting an external pointer: [type=environment]. at shutdown. That happens on current master for any dummy object whose finalizer runs and is independent of this change; it could be addressed separately by having run_finalizer() (or CppObject__finalize) skip the dummy instance.

Verification:

• The repro from calling a method on an uninitialized module object aborts the R session #1495 aborts (SIGABRT, exit code 134) against current master and prints caught R error: C++ object not initialized. (Missing default constructor?) with exit code 0 against a build of this branch.
• R CMD check with RunAllRcppTests=yes passes on R 4.6.1.

Checklist

• Code compiles correctly
• R CMD check still passes all tests (checked locally with RunAllRcppTests=yes on R 4.6.1; --no-manual --no-vignettes, so the only flagged items were the two vignette-packaging warnings from the local --no-build-vignettes build)
• Preferably, new tests were added which fail without the change (without the fix, the new test terminates the test process)
• Document the changes by file in ChangeLog

[eddelbuettel]

LGTM.

Interesting that 'merely' wrapping the standard entry / exit of generated functions does the trick.