Fix false file error reporting

See merge request !514

Lua: Fix some errors not working properly.

See merge request !510

Use M_FileError to return the proper error description, or "end-of-file".

Since the enum type is implementation-defined, and could be either signed or unsigned.

Compare the file name only for real

See merge request !511

CL_SendRequestFile prepends the path for each file.

Update version to 2.1.25

See merge request STJr/SRB2!507

Fix an exploit where players could steal the final hash of a login

See merge request STJr/SRB2!508

PolyObject crush hotfix

See merge request !506

Use P_MobjInsidePolyobj instead of P_MobjTouchingPolyobj, so that you can be crushed by the polyobject's middle rather than just its edges

PolyObject fixes backport

See merge request !505

After looking at the FOF part of P_LineOpening for a while I now realise many of these variables aren't even necessary, so I removed them all.

(Naturally I did the same to the camera equivalent)

# Conflicts:
#	src/p_maputl.c

Edit a lot of the rest of the polyobject-related code in P_LineOpening to make more sense and be more optimised.

* If you collide with a line belonging to a polyobject, you should NEVER have to care about any FOFs that might be present in either sector of the linedef. This could lead to colliding with ghostly FOFs that aren't actually there or something dumb, if someone decided to give either of the polyobject's control sectors FOFs for some reason. We don't want that, obviously.
* Polyobjects without POF_CLIPPLANE apparently are supposed to have a top and bottom "physical" height of value INT32_MAX and _MIN respectively, according to P_CheckPosition ...let's be consistent with this.
* Finally, there is no more need for that back = front nonsense hack anymore with my changes made.

# Conflicts:
#	src/p_maputl.c

P_LineOpening: set int32 max/min as defaults for opentop, openbottom etc if a linedef you touched belongs to a polyobjetc. the only thing that really matters in this scenario is the polyobject itself after all!

# Conflicts:
#	src/p_maputl.c

Modify P_CheckSector with a modified version of Sal's attempted proper fix for polyobjects crushing, so that we only need to check the polyobject's control sector directly in the waypoints code.

This time I've definitely fixed that teleport to ground issue I'm pretty sure, I don't get it in my tests at least.

This reverts commit 749c2753.