General tidy and restrict cffi < 1.17 due to static condensation demo issue

[jhale]

Not totally clear why but there is an issue with the static condensation demo, numba and a new version of cffi. @michalhabera you can reproduce this by branching and removing the cffi<1.17 restriction from pyproject.toml.

This PR also tidies up all of the CI pipelines; I would like to ask again, please do not put pip install x y z into pipelines. Please use the optional dependencies feature within pyproject.toml.

I have also split apart cffi registering Numba's complex types and creating cffi PETSc wrappers - you can now use cffi with numba complex numbers without having PETSc installed. This was causing a test failure.

[jhale]

See also:

#3340

[michalhabera]

I can reproduce. The issue seems to come from some changes in CFFI, so the global FFI builder object, i.e. cffi.FFI() does not recognize float _Complex and double _Complex types, call to ffi.typeof("double _Complex") gives

Traceback (most recent call last):
  File "/dolfinx-env/lib/python3.12/site-packages/cffi/model.py", line 58, in get_cached_btype
    BType = ffi._cached_btypes[self]
            ~~~~~~~~~~~~~~~~~~^^^^^^
KeyError: <_cffi_double_complex_t>

What seems to work is if we call typeof("double _Complex") on the FFI of the out-of-line compiled module. I'll prepare a PR.

[michalhabera]

Fix will come with python-cffi/cffi#111, I'd suggest we wait with pinned cffi==1.16.0 in the meantime.