PaAlsa_EnableRealtimeScheduling fix

[daschuer]

This fixes a crash I have uncounted during my testes with different Portaudio version.
Without this patch, the PaAlsa_EnableRealtimeScheduling writes a random byte in the JACK case and if the PaAlsaStream structure changes.

[uklotzde]

What's the precise reason for the crash? Are we passing a structure layout that has been determined at compile time to a dynamically loaded library that might use a different layout internally?

[daschuer]

Yes, that is one reason, that only happens during development.
The more serious reason is that in case of Jack, Pa casts a PaAlsaStream structure to a PaJackStream structure with different data and length.

[Holzhaus]

There's a merge conflict. Also, I can try to do a code review but I have no idea how to test this properly.

[daschuer]

@Holzhaus There is not much to test. If it does not crash it is fine.
The PR basically replaces error prone code with missing definition in the portaudio API.

[Holzhaus]

There is not much to test. If it does not crash it is fine.

It doesn't crash. I can't really judge the code though.

[daschuer]

Do you have questions?

[uklotzde]

I still don't completely understand how this works and why it is safe. Do you have any references or links that we could add to the code?

[daschuer]

I have just realized the the pa_linux_alsa.h is now part of the distribution.
So we have now no hack at all.

Here is the reference link. I don't think there is any need to put it into the code, now we have the header.
http://portaudio.com/docs/v19-doxydocs-dev/pa__linux__alsa_8h.html

[uklotzde]

LGTM, awaiting CI results.

[uklotzde]

Check the CI failures. All seem to be unrelated. LGTM