Experimental Fixes to the Mapgen

[Saalvage]

I find it an interesting study to fix the mapgen algorithm so it never throws with differing map sizes, while maintaining exact consistency for all existing, working configurations (sizes + seeds).

To make sure that none of the changes chance existing seeds we should generate a large corpus of generated maps we can compare the fixed algorithm's output to.

A few things to consider:

• Different map sizes take differently sized pieces of their provided pie, this gives rise to the consideration of what the actual size returned by the mapgen should be. Shrinking the array to fit the map seems like a reasonable solution. An alternative would be taking the maximum available (the internal array's size).
• Configurations that utilize the horizontal edges end up corrupted, same goes for configurations with very little vertical space (it should be checked if/how this behavior differs from the original implementation, even if it's erroneous).

Cute lil' broken map