Fix x0 and possible memory leak

[nbelakovski]

In #148 the ability of the user to specify x0 from the C interface was removed. This PR fixes that. I noted this at the time in a comment in 488abda but I suppose you missed it.

In the issue you objected to setting result->x to x0, but there is no other reasonable way to provide x0 to the Fortran code. In the issue you mentioned that result->x was set to a reasonable value even though no solver was selected, but the correct solution to that would be to have prima_check_problem return a non-zero exit code when it doesn't select a solver, as I mentioned in the issue.

Separately, I discuss this in my comment on the commit linked above as well, the modifications from that issue led to a possible memory leak. If prima_init_result returned 0 but prima_check_problem did not, we would return to the user with a non-zero exit code and also with allocated memory that the user would be expected to free. Of course it's not appropriate to both return a non-zero exit code and expect the user to clean up after us. The easiest way to solve this problem is to just check the problem before initializing the result.

[zaikunzhang]

Thank you @nbelakovski . I see the problem. I was confused by the original comment.

I will merge it for the moment, but still have related changes to make.

Thanks.