Realised the potential for out-of-bounds memory accessing, so put some limits in place to prevent that from happening.

(skin->availability is a UINT8, so it'll never be negative)

Setup:
* S_SKIN needs an "availability" line. Set to 0 to have always available, otherwise is assumed to be UINT8 unlockable number. If it's a valid unlockable, its name will be copied over with the character's realname.
* Define an unlockable via SOC or hardcode. If you don't want it visible on the Secrets menu, just set its type to None. Everything else is as normal - you can set the conditionset to whatever, objective, height, nochecklist, nocecho...

Also, new default for revitem to match player.dta, and an updated Force Skin netgame option name to reflect the fact it's string based now.

Meanwhile, a sneaky sneaky for hiding hidden characters further.

Also, I guess CA_SWIM isn't forced into running on water anymore.

* radius - sets the player's radius for that skin.
* height - sets the player's normal height for that skin.
* spinheight - sets the player's spinheight for that skin.
* shieldscale - see http://i.imgur.com/BQ5DhKC.png for justification

Rudimentary SOC shuffling if you add a character with the same name as a previous character to prevent character select from breaking.

More clearly defined handling for replacing character select images. Specifically takes into account whether the hidden character was added in this file or a seperate one.

Also, removed some commented out code regarding character selects, and removed a useless case for R_SkinAvailable().

* support hidden characters
* take the skin->name, not the number
* display the skin->name in the menu

Also, minor tweaks to other things.

R_SkinUnlock defines the circumstances under which a skin is available. For simplicty's sake, I've currently bound it to an S_SKIN variable so I can toggle it easily, but it WILL be replaced with a hook into the savegame system at some point.
* Currently has three tiers of unlock - freebie (forceskin or modeattacking via a loaded replay), Ringslinger Only, and SP/Coop and Ringslinger.
* I don't know anything about netcode so I basically decided to make R_SkinUnlock relevant only under local circumstances, try as hard as possible to stop bad skin info from getting sent to the server, and then admit defeat once the server has the information. If this is a bad choice, please discipline me and show me how to fix it.
* Character Select now checks for whether the character is hidden or not on menu load and does/undoes it based on that info, but will never touch one disabled via SOC. I also used this opportunity to optimise, checking for/filling out charsel pictures instead of doing it later. (It now also includes special casing for a select screen with zero characters!)
* Mode Attack now hides hidden characters in its character select based on SP rules.

Things that still need to be done:
* ForceSkin_OnChange. Is there a graceful way to handle this?
* No obvious skin name conflicts. Add a salt to the names of hidden skins, and then remove it when they're unhidden?
* The gap between Knuckles' skin number and the first custom character anybody adds will be way too obvious. A seperate hidden skin numbering system? Start at 32 and count up from there? There's a few ways...

Also, bugfix for something my optimisation introduced.

* If a character select character image is not set, don't iterate every tic - iterate on first image get and then save to the struct.
* A character select screen with only two characters now has special case handling.
* A memory leak in the making has been plugged. (specifically, picname not being Z_Free'd if the loop fails to do so)
* Logic/operation simplification.

Also, some typo corrections and clarity case movements of stuff in other files I've been looking at.

* SF_NOJUMPSPIN - Player's height is full whilst jumping, SPR2_JUMP defaults to SPR2_SPNG instead of SPR2_SPIN, and the player goes into fall frames if they start moving downwards or use their ability.
* PA_JUMP - for jumping (upwards in the case of SF_NOJUMPSPIN.
* SF_NOJUMPDAMAGE - Ala rosy.wad, don't damage enemies, etc when jumping into them.
* SF_STOMPDAMAGE - Just for fun. Ala in Mario, always damage enemies when you land on top of them (your gravity reference, not theirs).
* SF_MARIODAMAGE - SF_NOJUMPDAMAGE|SF_STOMPDAMAGE is reasonably accurate to the Mario games, and might as well be surfaced as such.

Also, a minor change:

* Instead of not spawning the revitem if your SPR2_ is SPR2_DASH, don't spawn it if it's set to 0. This requires the player.dta I uploaded a couple days ago to behave as it was previously.
* Don't get stuck in spindash frames if your maxdash is 0, and don't flash rolling frames if you're on goop.

Significant improvements to the Character select screen in preperation of implementing unlockable characters.

* Characters disabled through SOC are outright removed. No awkward gap in scrolling - no hint they were ever there in the first place.
* Vertical loop - the character select images are visually continuous. No matter where you are in the chain, you'll always see a hint of the character above or below your current selection in the chain.
* Smooth scrolling - Moto and Prime showed me a gfy from back during 2.1 development where it was super smooth. I didn't make it as slow as that one, but the smoothness was easy to add and the reason it was removed previously - gaps in the character select leading to varying speeds - is no longer relevant.

Fix for rendering sprites in skyboxes

This branch fixes the visual glitches one sees with sprites for objects within skyboxes, typically with parts of or all of some sprites disappearing from view even though they shouldn't be.

See merge request !84

OpenGL slope fixes again

This branch just adds the relevant code for OpenGL to properly check slopes regarding "closed door" segs (those between two sectors that cannot be crossed normally) and "window" segs (those between two sectors of differing plane heights that you CAN cross, which would probably display top or bottom textures), so you don't get HOMs when the the slopes create a closed door even though the normal sector heights wouldn't or something.

(if you couldn't understand that, my slopes test map shows what I mean to the right of the player start: https://dl.dropboxusercontent.com/u/25409000/2.1/mi-slopetest.wad)

See merge request !81

Basically this makes sure numwadfiles is updated before loading the SOC/Lua scripts, so if a Lua script calls COM_BufInsertText with the contents "addfile scr_mysticrealm.wad" it can't overwrite the last written wadfile slot! Not that COM_BufInsertText really should be used like that to begin with

Reviewed by @RedEnchilada