Split nmod_poly_factor out from nmod_poly
Remove arb from setup.py

Added a skeleton pyproject.toml
added a simple setuptools setup.py

THere is no good replacement to use instead of numpy.distutils to get
the default library and include paths.  Any suggestions would be
welcome.

The arb lib is now part of flint so all paths are prefixed with flint/

added missing fmpz_mpoly_sort_terms to _flint.pxd
updated fmpz_mpoly_context to include names and other term orders
Fixed up arithmetic to use python 3 protocol and removed automatic
conversion.

Made fmpz_mpoly_ctx return a string instead of bytes.

I hope these work for CI.  I currently don't have flint 3.0 or whatever
the matching arb lib is.

Some slight adjustments to flint_mpoly_context and fmpz_mpoly needed.
It now works on my machine.

Also added tests, and a few __iop__ methods to fmpz_mpoly

Yes, this does not compile.  I need to go off on another branch to
experiment with solutions.

Added infrastructure for fmpq_mpoly
fixed up new import system

Forgot to add it to last commit

Also respliced fmpz_mpoly and fmpq_mpoly back into doctests.

Added submodule to setup
Added rational function class to flint_base
Added interface to flintlib
Added skeleton fmpz_mpoly_q to types

Added missing fmpz_clear for exponents in creating fmpz_mpoly from dictionary
Fixed spacing in converting fmpq_mpoly to string

also basic ops on fmpz_mpoly_q

Also added method to get context from a fmpq_mpoly so the tests could check that
it was correct.

Moves ctx lookup into something based on the ctx class rather than aptly named functions.

More to just get myself acquainted with the code base than anything else.

This caused too many headaches during testing. Since the order of variables within a context is not stable during joint
context construction, it's simply one giant walking gotcha.

Add getitem and setitem, add derivative and integral, and degree and total_degree.  Add IncompatibleContextError. Fix
behaviour of div functions to align with new domain errors. Add sqrt.

I'm experiencing a really weird issue with test coverage atm, it's influencing the out come of the getitem tests?

TODO: Remove coercion in operators, copy work over to fmpq_mpoly

This reverts commit fa07adf.

Line 369 is being run I swear, coverage just isn't picking it up