[PATCH 0/1] MR3458: ddraw/tests: Fix overflow bugs in compare_float.

[Jeffrey Smith (＠whydoubt)]

1. Subtracting integers may result in an overflow or underflow.
2. Right at the 'edge' of underflowing, the result of subtraction may be
   `INT_MIN`, and the call to `abs()` will also result in `INT_MIN`.

This fix accounts for all of these conditions.

EXAMPLES:
1. can be encountered by comparing 2.0 and -2.0
2. can be encountered by comparing -2.0 and 2.0

NOTE:
There are 14 more instances of `compare_float` across several modules.
I would like to make sure the logic is sound with this MR, then I can take on the rest.