…UDE]

Review feedback from #7558:

- sqlglot/__init__.py: drop the *__mypyc*.so bootstrap preloader; legacy
  hash-named .so layout doesn't happen with separate=True.
- sqlglot/optimizer/__init__.py: rename _t to t, tighten the docstring.

Drop py3.9 support for sqlglot[c]:

sqlglot-mypy 1.20+ doesn't ship a py3.9 wheel and the older 1.19.x line
(which has a universal py3-none-any wheel) predates the separate=True
fixes, so py3.9 CI was crashing with "vtable not computed yet" while
trying to build sqlglotc.

  - setup.py: gate the [c] / [rs] extras on python_version >= 3.10, so
    `pip install sqlglot[c]` on 3.9 is a no-op (you get pure-Python sqlglot).
  - sqlglotc/pyproject.toml: bump requires-python to 3.10, pin
    sqlglot-mypy >= 1.20.0 in build + dev requires.
  - Makefile: install-devc / install-devc-release skip the sqlglotc
    build on 3.9 instead of failing.

sqlglot itself still supports py3.9; only the optional compiled extension
requires 3.10+.

…UDE]

Review feedback from #7558:

- sqlglot/__init__.py: drop the *__mypyc*.so bootstrap preloader; legacy
  hash-named .so layout doesn't happen with separate=True.
- sqlglot/optimizer/__init__.py: rename _t to t, tighten the docstring.

Drop py3.9 support for sqlglot[c]:

sqlglot-mypy 1.20+ doesn't ship a py3.9 wheel and the older 1.19.x line
(which has a universal py3-none-any wheel) predates the separate=True
fixes, so py3.9 CI was crashing with "vtable not computed yet" while
trying to build sqlglotc.

  - setup.py: gate the [c] / [rs] extras on python_version >= 3.10, so
    `pip install sqlglot[c]` on 3.9 is a no-op (you get pure-Python sqlglot).
  - sqlglotc/pyproject.toml: bump requires-python to 3.10, pin
    sqlglot-mypy >= 1.20.0 in build + dev requires.
  - Makefile: install-devc / install-devc-release skip the sqlglotc
    build on 3.9 instead of failing.

sqlglot itself still supports py3.9; only the optional compiled extension
requires 3.10+.

…fy [CLAUDE]

Temporary debug aid for #7558: integration tests are producing a
~530KB wheel that's missing all shared-lib .so files, but local builds
produce the full 18MB wheel. Add stderr prints so pip's failure output
shows what python/mypyc/source-dir setup.py actually saw, plus a hard
exit if mypycify returns zero shared-lib Extensions, so we get a clear
error instead of a silently-broken wheel.

Will revert once we've identified the root cause.

…fy [CLAUDE]

Temporary debug aid for #7558: integration tests are producing a
~530KB wheel that's missing all shared-lib .so files, but local builds
produce the full 18MB wheel. Add stderr prints so pip's failure output
shows what python/mypyc/source-dir setup.py actually saw, plus a hard
exit if mypycify returns zero shared-lib Extensions, so we get a clear
error instead of a silently-broken wheel.

Will revert once we've identified the root cause.

…fy [CLAUDE]

Temporary debug aid for #7558: integration tests are producing a
~530KB wheel that's missing all shared-lib .so files, but local builds
produce the full 18MB wheel. Add stderr prints so pip's failure output
shows what python/mypyc/source-dir setup.py actually saw, plus a hard
exit if mypycify returns zero shared-lib Extensions, so we get a clear
error instead of a silently-broken wheel.

Will revert once we've identified the root cause.

…fy [CLAUDE]

Temporary debug aid for #7558: integration tests are producing a
~530KB wheel that's missing all shared-lib .so files, but local builds
produce the full 18MB wheel. Add stderr prints so pip's failure output
shows what python/mypyc/source-dir setup.py actually saw, plus a hard
exit if mypycify returns zero shared-lib Extensions, so we get a clear
error instead of a silently-broken wheel.

Will revert once we've identified the root cause.