[freeciv-ai] (PR#2477) Improved Auto-Explore

["Guest via RT" <rt@xxxxxxxxxxxxxx>]

I fiddled with the values, and the result seems better. I made far away
tiles less desirable, and changed the relative value of same/different
tiles.

Judging merit of the explorer code by comparing civilization score is
too crude, the noise is immense. With the current values I have never
seen a situation where I think it did "the wrong thing", so I believe
the current values are as close to optimal as could be expected.

Also, I made equally desirable tiles be selected in a more fair manner,
by adding a very small random value to the desirable score. Previously
if a tile was equally desirable to the previous best tile, there would
be a probability of 0.5 of using the new tile, and 0.5 of keeping the
old. So if we have, for example, 3 equally desirable tiles, they will
have probabilities of 0.25, 0.25, and 0.5. With 4, they are 0.125,
0.125, 0.25, and 0.5. With a random value added to each, with any number
they are close to being randomly chosen.

explorer8.diff
Description: explorer8.diff