calcfc_chebyqud and calcfc_hexagon in Python return values close to what is returned by Fortran,
but not exactly the same and the difference is outside of machine precision.

For chebyquad the fortran version iterates 387 times within cobylb, whereas
the Python version iterates 564 times. The results are as follows

Fortran:
x(1) = 0.06687201625695266
x(2) = 0.28872054433564054
x(3) = 0.36668681233017669
x(4) = 0.63330363432938008
x(5) = 0.71126615229302259
x(6) = 0.93312277720964698
f = 3.9135366518694255e-10

Python:
x[0] = 0.06688196439284098
x[1] = 0.2887610493288529
x[2] = 0.36668226015206873
x[3] = 0.6333320849097046
x[4] = 0.7112583722245663
x[5] = 0.9331254078707577
f = 3.8259601181730434e-10

The first position in x differs by 9.948135888e-6, so it's further away than machine precision, but
it's possible that the difference is the result of accumulation of machine precision errors.

Implemented savehist and modified the return values to be more usable.

I've gone through cobyla.py and it seems pretty close to the original.

Remaining work is to implement retmsg and to add in all pre/post conditions
and retrofit existing pre/post conditions to match the style guide.

Then we're ready for a deeper conversation on whether any of this is useful XD