add basic support for SCO packets over USB

[barbibulle]

Still WIP

[zxzxwu]

@barbibulle This PR doesn't work with Intel controllers - but I asked Gemini to fix it #930, and now I can receive SCO but only with CVSD, but I guess transparent packets may work.

It seems the major difference is that alt setting must be set dynamically? And also input packets must be properly segmented.

[barbibulle]

@zxzxwu Would it be possible to rebase #930 on this commit, so that I could easily see what the gemini fix does? (maybe there's a way to do that easily from the GitHub interface, but I don't see it)

[zxzxwu]

Sure, I rebased it and pushed to #930.

[barbibulle]

Sure, I rebased it and pushed to #930.

Thanks.
I took a quick look at the gemini fix. It's interesting. There's one good element in there, but also some funny mistakes :-) In fact, it works by accident, because of a bug that it created but that ends up calling something with a value that works. Here's what's going on (based on a quick experiment with an Intel controller, I need to do some more experiementation with it). With my Intel controller, Alt Setting 6 does not work. When specifying +sco=0, the usb transport will auto-select the setting with the larget packet size, which is this one. The reason it works with the gemini change is that it created a set_sco_config to change the alt setting. It created that function with the alternate setting as the second parameter. But then it gets confused, having another function async def set_sco_config(self, active: bool, air_mode: int) which uses air_mode as its second parameter. In the end, it ends up using neither, because in the body of set_sco_config, it will set the alternate setting obtained from earlier from self.sco_alt_setting =sco_interface.getAlternateSetting()...
So, skipping all that spaghetti, all you need for the Intel controller to return SCO packets is select an alternate setting that's not 6. I tried it, using usb:0+sco=1 instead of usb:0+sco=0, and I get SCO packets.
Gemini did the right thing in adding some code to split the buffers into smaller packets, and using multi-packet transactions. That part I will incorporate in my PR.
The audio sending look in the sample app is good too, I had not had a chance to do that yet. So that's cool.

[zxzxwu]

Checking on BE200, CVSD is also support. mSBC codec is not supported (despite claimed in supported codec command), but transparent is supported, so we can encode and decode them in host.

[zxzxwu]

@barbibulle Thanks! What's your plan for this PR? We would like to use BE200 for some purposes especially on SCO and ISO.

[zxzxwu]

Can this be a daemon?

[barbibulle]

I was thinking of just adding the outgoing packet splitting and then try to merge it, then add the multi-packet input parsing as a second PR (this will address an issue reported earlier). For Intel controllers, we can either document that some alternate settings may not work, and that users should select a setting other than 6 manually, or we can make the auto-selection skip config 6, but that may not be ideal, since it might not be broken on all controllers, and since it has a larger packet size, may be more efficient. What do you think?

[zxzxwu]

Maybe we can allow passing a spec to select alt settings with a filter? For example:

• sco=0 : For auto selection
• sco=N : For fixed selection
• sco=N,M : For auto selection from either N or M
• sco=N-M : For auto selection from N to M