msolve hangs/segfaults when non-zero coefficient reduces to zero over finite field

[wegank]

In the following test:

msolve/input_files/input-overflow-16.ms

Lines 1 to 4 in 6c9d63e

	y,x
	65519
	x^2+128333*x*y-12*y,
	y^2-x

The aim is clearly to test whether 128333 can be normalized. However, if we replace 128333 by any multiple of 65519:

y,x 
65519 
x^2+131038*x*y-12*y, 
y^2-x 

Then msolve hangs. Of course, the normalization introduces a zero coefficient...

Similarly, for the following input, msolve segfaults.

x
7
7

Also, I don't see why one should do

msolve/src/msolve/iofiles.c

Lines 749 to 751 in 6c9d63e

	while (iv_tmp < 0) {
	iv_tmp += field_char; //MS change int -> long int
	}

while we have modulo operations in C?