Carla plugin fails to build with 'develop' branch of Carla

[PhysSong]

As of falkTX/Carla@d4a0292, Carla plugin fails to build. The API-breaking commit changed the type of inBuffer parameter of NativePluginDescriptor::process from float** to const float**.

The conversion from float** to const float** can't be done implicitly, while float**-to-const float* const* conversion can be implicit(in C++, not in C).
This means if we want to fix the build for newer versions of Carla, we must drop support for old versions(which Mac builds use) since the only way to convert pointers between float** and const float** in C++ is explicit const_casts.

@falkTX

[falkTX]

There are some other changes actually, that make it ABI incompatible between 2.0 and future 2.1
I took the chance to cleanup the API since backwards compatibility was going to be broken anyway.
This is all happening on the develop branch, for which there might still be more changes.
I will send a PR-fix to LMMS when I get finalize tests and decisions.

This is not yet something that LMMS needs to be concerned about, I think.
Though if building packages for Linux still uses KXStudio repos for its carla-git (I guess this is how you noticed the breakage?) then simply change the package name to carla.
I am releasing v2.0 as stable in the next few days, the changes in the repos reflect that.

See also https://kx.studio/News/?action=view&url=changes-in-kxstudio-repos-regarding-carla-and-jack2

[PhysSong]

As of 032c324, Travis-CI builds use carla. Please let us know when you make a final decision.

[falkTX]

As of falkTX/Carla@0514349 Carla 2.1 plugin ABI is now backwards compatible with 2.0 again.
So both versions are fine, though I recommend to keep using the carla stable version without git/develop[ment] stuff.

Note there might be some minor compiler warnings building against 2.1 of carla, since new enum values got added. but they are all optional

[tresf]

As of falkTX/Carla@0514349 Carla 2.1 plugin ABI is now backwards compatible with 2.0 again.

@falkTX working on #4813 I encountered the same exact issue mentioned above using the tip of the develop branch, so I believe this statement to be incorrect, no?

The temporary fix for #4813 was to checkout the submodule at revision v2.0.0, which is currently working just fine.

[falkTX]

ah I forgot this part. I will update Carla so that it matches old API.

[falkTX]

Should be fixed now, please confirm

[tresf]

Should be fixed now, please confirm

Seems to work just fine, thanks!