When using FormatText() several times in same function, returned string is static and so, the same is returned, resulting in failures on shader loading.

From the Clang documentation[1]:
> -x<language>, --language <arg>, --language=<arg>
> Treat subsequent input files as having type <language>

Follow the advice. Fixes #840.
[1]: https://clang.llvm.org/docs/ClangCommandLineReference.html

This reverts #841 commit 97160fd970.

[build] fixed GLFW compiler flag for OSX

Contributed by ProfJski

Now shader uses `fragTexCoord` that are the full screen texture coordinates normalized, instead of `gl_fragCoord`, the unnormalized screen coordinates

Update julia set example.

Also changed path to shader back to what it was originally.

Now dividing by the zoom instead of multiplying (in the shader), so zoom works as expected. Also zoom increase/decrease is now scaled depending on the current zoom.

Note that all meshes are loaded as a single one at this moment, loading should be improved!

This reverts commit 4773de26a5.

which adds -lwinmm at the wrong place. It should be in the raylib
linker flags, not GLFW's.

extra-cmake-modules 5.38.0a is no longer hosted at the specified
kernel.org link. Use the occasion to move on to 5.58.

Because newer ECM debian packages have the control.tar file
xz-compressed, we need to update dpkg as well on trusty.
For more information, see
https://bugs.launchpad.net/ubuntu/+source/dpkg/+bug/1730627

Fixes: e0e2346c22 ("NO SUPPORT_BUSY_WAIT_LOOP by default")

Note that building examples for web as they are (no code adaptation for web avoiding while loop) implies using the emterpreter... and that's very slow!

Redesigns, deletes and renames

Also noted authors propertly on contributed examples

Added julia set shader example.

Now it's required on Windows if not using a busy wait loop

Fixed DrawTextRecEx() selection when wordwrap is ON

Remove sqrt() from CheckCollisionSpheres()

Update CheckCollisionSpheres() to avoid sqrt

Square root calls are computationally expensive.  In this case, they can be avoided.  Instead of checking distance<RadA+RadB, check distance squared against (RadA+RadB) squared.  The dot product of Vector3Subtract(B,A) with itself gives distance squared, so I used this code instead of an element-by-element computation of distance squared.  The only downside is that your geometric code is very readable, whereas this is less so.