ising_matrix_readme

This version of the code uses a compiled C library that python calls to do 
the main math. It is a little more work, but runs faster (~6 times for a 50x50 matrix)

The process is this:
    1. write the C-library. This is done in ising_matrix.c
    2. compile the C-library. On OSX, this can be done with the following command:
        clang -shared -Wl,-install_name,ising_matrix.so -o ising_matrix.so -fPIC ising_matrix.c
    3. insert the ctypes module into main.py
        " from ctypes import * "
    4. using the C-library.
        1. load the library with a command like:
        " c_matrix = CDLL('./ising_matrix.so') "
        2. feed in "ctypes" into the C-library function calls, i.e. c_library.function_call( c_int(10), c_float(...))...
        3. python expects c-library function calls to return c_int's by default. If they don't, be sure to tell python
           with c_library.function.restype = <c_type>
        4. the way that the C-library is coded here, it requires that the function ".allocate(c_int(N))" is called first,
           otherwise it will seg fault when pretty much any other function is called.
    5. notes on the c-library coded here:
        1. There are two arrays. The array for the spins is just an array of booleans. This is True (or 1) for
           spins aligned with B, and False (or 0) for spins not aligned with B.
        2. The other array is pairs, which is a short int. It counts how many neighbors to each site are anti-aligned.
           e.g., a site with spin up adjacent to three sites that where spin down (and the last adjacent site, of course
           being up) would have a value of 3.
        3. The use of booleans for the spin matrix allows for a small memory footprint and use of bitwise operators
           to check spin alignment. This did, however, require that the forms used to calculate E, M, and deltaE
           to look different. However, the calculated values for E and M are identical to those if -1 and 1 were used instead of 
           the booleans.

    6.  Finally, the default random number generator from the C library is not
        ideal. There are better, and faster, RNG's. An obvious improvement would be
        to add a Marsenne Twister RNG into the code instead of using the
        rand()/MAX_RANDOM. The one from
        http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/TINYMT/index.html works
        well. There is no reason, besides time spent, that it is not already
        implemented.

For an brief introduction on using ctypes, see <<http://book.pythontips.com/en/latest/python_c_extension.html>>