Fix the usage of udev and dbus with SDL joystick input driver

[Nintorch]

In the SDL joystick input PR I decided to not compile it with libudev because it was a bit too late to make new changes.
Yesterday I found out that SDL actually uses dynamic loading of libudev and uses the <libudev.h> only for structs, typedefs, etc. in the thirdparty/sdl/core/linux/SDL_udev.c file, so I decided to fix the compilation of libudev to use the header file that was already bundled with Godot's source code (thirdparty/linuxbsd_headers/udev/libudev.h) and in another file (thirdparty/sdl/hidapi/linux/hid.c) where SDL actually does use functions from said header, I redirect the calls to the struct where they're loaded dynamically (the one in SDL_udev.c). And I removed the usage of SOWRAP_ENABLED in the detect.py file where UDEV_ENABLED is defined, because SDL loads the library dynamically anyway and we probably can't/shouldn't change the way it works.
I tested the PR with WSL and it didn't seem to crash or break the controller support (my Dualshock 4 and a cheap Sega Mega Drive-like controller work fine just like before), so I assume it works 😅 (EDIT: need testing again)

EDIT: also made dbus load "statically" too (i.e. pull its functions from Godot binaries just like udev without SDL_UDEV_DYNAMIC)

TODO:

• SDL can load the library "statically" if null is passed to dlsym() (i.e. search for the symbols inside of Godot binaries instead of a separate library file, I think?), so we can check for SOWRAP_ENABLED here
• Do the same for dbus

[Calinou]

Tested locally on Linux, it works as expected (including hotplugging). Code looks good to me.

Tested controllers:

• DualSense
• DualSense Edge
• Switch Pro Controller
• Xbox Elite Series 2 controller
• Nyxi Warrior

[Nintorch]

Thank you for your reviews, Calinou and bruvzg!

[Repiteo]

Thanks!

[Nintorch]

Thank you so much! :D