Slight cluster missile rework

[Alexandre425]

Slightly reworked cluster missiles so that the bomblets are properly modeled. It assumes 7 are packed in parallel per layer, with up to 4 layers, resulting in a maximum of 28 bomblets. The bomblet caliber and length is changed to reflect this. For any questions refer to https://en.wikipedia.org/wiki/Circle_packing_in_a_circle

The random spread is also significantly simplified by using a uniform angle distribution, randomizing only the distance from the center. It results in almost the same spread, only slightly more uniform:

Removed the relationship between bomblet spread and global inaccuracy, as it seemed a bit nonsensical (you essentially want the bomblets to be "inaccurate").

Physics-wise, it also makes more sense, with the bomblets gaining a velocity perpendicular to the deployer which is independent of it's original velocity. They now inherit the original velocity plus the separation velocity, instead of changing vector's the direction.

[Alexandre425]

Currently the bomblets for some bombs are very long and thin compared to conventional ones. Due to the way HEAT is coded, this gives them very light but extremely fast penetrators, with too high of a penetration.

I plan on reworking HEAT so the generated warheads make a bit more sense, and then I'll reopen this once I have tested and balanced it.