Log Message: 
Fixed conceptual error in compare_vec_solution. This should fix bug
#670. In my opinion the entire concept of vec_solution_cmp should be
reconsidered. In solving an underdetermined linear equation of the form
Axb=0 it seems to me that the solutions answer in the form:
x= a +bt+cu+ds where a,b,c,d are vectors should simply be evaluated
to see if it satisfies Axb=0 for 5 or six values of a,b,c,d 
checking the solution should use a
vector valued version of fun_cmp.
As it is, the student's coefficients for a,b,c,d are compared with the
instructors to see if they span the same space. This is quite a bit more
complicated  and indeed the method came up with the wrong answer.
I believe I have the method corrected, but I would suggest that this
answer evaluator be replaced with one which operates more
directly and is therefore easier to maintain. Am I missing something in
this analysis? Has someone else created answer evaluators for this type
of problem?
 Mike
