Do a full rewrite of the MOI wrapper. (#22)

There are still a few outstanding TODO items, including one serious segfault in
linear 8b.

The HiGHS API has a few quirks.
 - It is very talky.
 - There is no way to get the error message from a return code.
 - The duality convention for rows is different from columns by a factor of -1.
 - There is no way to change constraint coefficients.
 - There is no way to get the list of options, or the type they expect.
 - It would be useful to have Highs_changeColsCostByRange
 - There is no support for primal/dual rays
 - No partial solutions are returned if the solver terminates early.

README.md

@@ -1,6 +1,3 @@
-**Warning: The MathOptInterface wrapper is still under development. It is missing**
-**a lot of features.**
-
 # HiGHS.jl
 
 [![Build Status](https://github.com/jump-dev/HiGHS.jl/workflows/CI/badge.svg?branch=master)](https://github.com/jump-dev/HiGHS.jl/actions?query=workflow%3ACI)