Support for free-threaded CPython (3.14t or newer) #503

This issue is intended to track the efforts of adding support for the free-threaded build for CPython (3.14t and potentially 3.13t). The Python free-threading guide contains some helpful advice on approaching this task. Based on what has been done for pybind11, I think the following steps make sense:

• Add mechanism to set the Py_MOD_GIL_NOT_USED. I have a patch that does something similar to what pybind11 does: add a variadic macro with an optional flag so extensions can opt-in to free-threaded support
• Replace borrow based APIs with strong reference ones (e.g. PyDict_GetItemRef)
• Use a mutex to protect global state (e.g. the registry)
• Fix references to ob_refcnt, don't use Py_REFCNT() as an l-value.
• Modify tests so we can run them in parallel, run with the TSAN build to find races
• Fixes to "faber" to work with new versions of Python. The most recent release of faber fixes most things, I believe.

I have code that does most of this, I will create a PR. I'm still working on some crash bugs when running with 3.14t. If you run the Boost.Python tests with a single thread, they pass. So that's encouraging.

Related issues:

• compile with python 3.13t (nogil) #458

Related PRs:

• Support for free-threaded build of CPython #504