stubtest: don't report re-exported submodules as missing from the stub

[ShirGanon]

Fixes #21328

When a package's __init__ binds a submodule as an attribute at runtime — e.g. via from .submod import * (or simply from . import submod) — and lists it in __all__, stubtest reported the submodule as missing from the stub, even though the submodule has its own stub file and is verified separately. This produced two spurious diagnostics per submodule:

error: pkg.submod is not present in stub
error: pkg.__all__ names exported from the stub do not correspond to the names exported at runtime
  Names exported at runtime but not in the stub: ['submod']

Python's import system binds an imported submodule as an attribute of its parent package, so listing such a submodule (which has its own stub) in the package's __all__ is correct, and stubtest shouldn't flag it.

Changes

• Added _get_stub_submodules() to find the direct submodules of a package that have their own stub (present in _all_stubs).
• In verify_mypyfile, skip reporting a runtime submodule as "not present in stub" when it has its own stub file (it's verified separately).
• In _verify_exported_names, exclude such submodules from the runtime-vs-stub __all__ comparison.

Submodules without a corresponding stub are still reported as before, so this only suppresses genuine false positives.

Testing

• Added test_reexported_submodule_in_all in mypy/test/teststubtest.py, which builds a package whose __init__ imports a submodule and lists it in __all__, and asserts stubtest no longer reports it as missing.
• Confirmed the new test fails without the stubtest.py change (it reports submod is not present in stub) and passes with it.
• Full mypy/test/teststubtest.py suite passes (67 passed), no regressions.
• ruff check, black, and mypy self-check on mypy/stubtest.py are clean.

_{Disclosure: drafted with Claude Code; I reviewed, understand, and tested the change locally.}