Ensure m3d faces in non-decreasing materialid sequence

[orcmid]

This modification replaces the expensive qsort protection with an insertion sort that is near-instantaneous in the expected ordered case.

The changes desk-check and they compile. It is not confirmed in a running application and I am guessing that I have satisfied raylib coding style.

PS: This is modeled on Donald Knuth's Algorithm 5.2.1S (Straight insertion sort) and is a refinement of that and the code @DaveH355 provided earlier in order to handle mostly-ordered faces as quickly as possible.

[orcmid]

I should add that this defensive approach is consistent with the steps taken a few lines below where this modification is inserted:

 // there should be only one material switch per material kind, but be bulletproof for non-optimal model files

[raysan5]

@orcmid Thank you very much for the review but as stated in the M3D specs, this sorting is not needed any more. I prefer to avoid extra code for non-allowed side cases. In any case I can replace the qsort() statement with this code but it should remain commented, only for reference.

[orcmid]

I think most folks are very reluctant to maintain a patched raylib, and only a small subset will even read this, so commented-out defenses don't seem particularly helpful to me.

I'm not going to argue. It's your project.

[raysan5]

As commented, I'm merging it for reference but keeping it commented. It is not required for as per the official M3D specs.

[raysan5]

Side note, raylib does not use the do { } while approach anywhere in the library.

[orcmid]

@raysan5

Side note, raylib does not use the do { } while approach anywhere in the library.

In this particular case it works perfectly along with rapid checking for already-sorted elements. I confess that I had to look it up to ensure I was using it correctly :).