libobs, win-wasapi: Enable channel routing for Audio Input Capture

[pkviet]

Description

On libobs side, this adds the capability for the audio resampler (relying on swresampler lib) to
set a matrix for routing of channels.
On win-wasapi side, this applies the new capability to enable channel selection.
This brings parity to wasapi input capture with coreaudio input capture on macOS.

Screenshots

Motivation and Context

One motivation for that feature are devices which separate stereo inputs.
So if a mike is connected to , say, Input 1 while nothing is connected to Input 2, wasapi usually treats them as
stereo pairs and when obs audio is set to Stereo, the captured Right channel will be silent.
With channel routing capability, it is now possible to duplicate the left channel to the right and have a consistent level.
The issue was brought to my attention by @Fenrirthviti

How Has This Been Tested?

Tested on windows 11 pro 25H2 w/ a logitech webcam + USB sound board.
Disabling the downmix cehckbox enables channel routing.
Swapping between devices resets the channel selections.
Changing the channel selections updates the captured channel.

Types of changes

• New feature (non-breaking change which adds functionality)

Checklist:

• My code has been run through clang-format.
• I have read the contributing document.
• My code is not on the master branch.
• The code has been tested.
• All commit messages are properly formatted and commits squashed where appropriate.
• I have included updates to all appropriate documentation.

[PatTheMav]

Looks fine to me overall, also tested this with an virtual 6-ch audio device and allowed me to map them as expected.

Haven't done in-depth checks whether the audio is correctly routed though.

[psyirius]

Just wondering,
Why don't we have speaker-layout selection like obs-asio has?

[psyirius]

Just built this pr from the current master.
Can't able to select channels in the channel comboboxes.

[pkviet]

Just built this pr from the current master. Can't able to select channels in the channel comboboxes.

yes there was a regression but it is fixed with last push

[pkviet]

Just wondering, Why don't we have speaker-layout selection like obs-asio has?

Because I want to avoid an uncontrolled upmix or downmix.
In obs-asio , if the input audio format differs from obs format (set in Settings > Audio), there is automatically an swresampler called in the audio pipeline which will make the conversion. The channel weights are sane only for the most common conversions (xxx ==> stereo, stereo ===> xxx ) as stated explicitly by the swresampler API. There can be surprising defaults with odd channel volumes otherwise.
There is no conversion though if the input speaker format is the same that is set in obs settings.
Channel routing is imo an advanced feature for people who want fine grained audio control; so the uncontrolled upmixer/downmixer insertion defeats that purpose and that is why I have avoided that here, very deliberately.