v0.6.0 missed the MIP index: mip-split.toml still names the removed multimer backends #103

v0.6.0 reached TestPyPI (pydevices and pydevices-desktop) and its GitHub Release, but the MIP index never took it. The mip sync rejected the tag:

mip-split.toml [multimer] names modules that no longer exist: librt, sdl2, threading, wasm, win32

Run: https://github.com/PyDevices/mip/actions/runs/36238913072

#101 (the timing redesign) removed those five backends, and mip-split.toml's [multimer] host-only list still names them. Nothing in this repo's CI reads mip-split.toml, so the unit tests stayed green and the error only showed up after the tag existed. The MIP index still serves 0.5.5 (pydevices-lock.json ref v0.5.5).

What to do: rewrite [multimer] host-only for the new modules, then cut a follow-up release (the tag can't be moved). A first guess from the module docstrings, which the redesign's author should confirm: host-only = _inputhook (CPython's PyOS_InputHook), _src_pending (CPython worker thread + Py_AddPendingCall), _src_signal (CPython and unix MicroPython), _src_native (the windows port's _timing helper), _src_wasm (browser). Boards keep _dispatch, _hostloop, _repl, _src_machine, _src_none, _src_asyncio, _ticks and the asyncio loaders. The comment above the list names the old modules too.

Worth adding: a unit test that every name in mip-split.toml exists under lib/, so this fails in the PR rather than after tagging.

Also, the Release Health row for pydevices 0.6.0 says MIP OK, but that only reflects the request job. The mip run itself failed.