Add function get_solution_count.

[jwg4]

No description provided.

[jwg4]

to fix #17

two functions to add:

• get_solution_count
• get_all_solutions

The first just returns an integer. The second returns an array of all the solutions. The second must be harder than the first. Let's do the first first.

• add the function to the Python module written in C
• add the function to dlx.c
• (probably) modify the basic algo code so that we can keep going after finding one solution
• add code in the Python wrapper to call get_solution_count
• add and run simple Python test
• add and run more complex Python tests
• C tests

[jwg4]

@parmentelat
Let me know if you have different opinions about the suggested ways of proceeding above.

I believe that based on my last push:

1. it will be possible to call the function get_solution_count from Python (in exact_cover_impl) module
2. it won't crash(?)
3. but it also won't give the right answer

so the next step is maybe to add a Python test

[jwg4]

added some tests based on a very simple ec problem

[jwg4]

the current function doesn't crash but clearly doesn't give the right answer. we have a good selection of tests.

[jwg4]

current function looks like it gives the right answer, next steps are to add C tests and clean up the C code a bit

I think it makes sense to merge this funcitonality when it's in a good state and then open a second PR for a function which returns an array of all the solutions.

[jwg4]

We need to figure out how the solution count should handle the following cases:

1. presence of blank rows. the number of solutions could be multiplied by 2^(number of blank rows) to represent all the possibilities of including them or not including them
2. presence of duplicate rows. the number of solutions could be multiplied by product(number of copies of each row in the solution) to represent all the possibilities

my guess is that the first will not happen (blank rows will be ignored) but that the second will happen (two rows the same means all the solutions that include one of them are counted twice). This also kind of seems like what one wants.

[jwg4]

this is published to pypi as 1.4.0a0

I think this probably can be merged with only documentation added, and then get_all_solutions which fills an array with solution vectors will be a separate PR. no point in cleaning up the C code until the second part has been done.

• mention how to use this in main README
• show how to use this in examples.md