Fix monster drops

I don't understand the 'uniques should always drop out-of-depth' comment. When I kill Wormtongue, and he drops an ego weapon, I usually want that weapon to be a dagger, main gauche, or rapier, and not an executioner's sword.

Also, I think eliminating DROP_GREAT is a mistake. Right now there are a few monsters that you go out of your way to kill. If you cut down on the size of Smaug's drop, or remove DROP_GREAT from Ar-Pharazon, they just become two more boring uniques.

In the current design, some monsters are really worth killing, and some are not. This was a conscious design decision, not an accident.

Take a look at Steamband. There is definitely not TMJ; in fact most monsters drop nothing at all. But there are still a few monsters that work like piñatas.

One thing I've started playing around with is the "good" drop frequency. Currently in the code, when an item of level N is requested (where N is usually the average of the dungeon level and monster level), the chance that the item will be good or better is 10+N, up to a maximum of 75. I think this should be boosted.

First of all, I think there's no reason to have a cap of 75. As it is, many deep monsters have the drop_good flag added. I've provisionally changed the probability of getting a good item in my copy to be 10+2N with no cap. I've also changed the probability of getting a great item; with the old code, the percent chance that a good item will also be great is N/2 (up to 20); in my copy, it's N/3 (up to 20). Furthermore, I've added a temporary hack to reduce all drops by about 1/3. This way, by dungeon level 45, no average equipment is created, and 20% of all equipment is great - whereas with the current formula at DL45, 45% of equipment created is average or worse, and a total of 11% is great.

Certainly, if we want to reduce drops, we need to make them better, and I think looking at the probability of getting a "good" item is a good place to start.

r760 has temporarily added back in some of the "kind_is_good" functionality, so that monsters with the "drop_good" flag are significantly more likely to drop "good" magic item types again.

r936, r940, r942, r943 all contribute to this. I think this is as done as it gets for 3.1; the TMJ problem has been scaled back massively and should start to appear ~dlev50 now. new-dlev100 is roughly equivalent in terms of drops to old-dlev65, so the curve has been made a lot less steep.