Allow WASAPI exclusive mode to work in more devices

[LAGonauta]

In all my devices exclusive mode wasn't working, and that is because GetCurrentPadding was always returning zero. This PR fixes in all my devices in my case, and there are no skipping or glitches.

sample-type=int16 is still required for my devices, but I am going to resolve it later.

[kcat]

Unless this value can change, it was already stored in mOutBufferSize, which is copied into the local variable buffer_len. It doesn't need to be queried each time. Also, since written is used to update mPadding, which is used for latency calculations, I'm not sure leaving it as the buffer size is a good idea (unless that is the actual amount of padding WASAPI holds for exclusive streams, i.e. the amount of audio still left to play before the new samples to be mixed).

[ThreeDeeJay]

Confirmed this fixed audio on my onboard Realtek device, though other devices are still having issues.
More details/context here: #682 (comment)

[kcat]

I ended up reworking stuff in the WASAPI backend, which created conflicts. But I may have also fixed some issues with exclusive mode. Are there still issues that need fixing?

[ThreeDeeJay]

@kcat I'm still getting no audio on my Realtek card (which is probably what most PC users have) https://www.diffchecker.com/gR1s8cfX/
My Creative X-Fi USB works fine with the same settings, though: https://www.diffchecker.com/THjvIkax/
Here's a comparison of the logs of both cards using WASAPI exclusive: https://www.diffchecker.com/RG6XGpcW/
I noticed that the Realtek changes period_size from 144 to 160, and still doesn't work. I also get no audio if I manually set it to 160 or even values over 1K.

[LAGonauta]

This should have been fixed in the rework.

[ThreeDeeJay]

Can confirm 👌
Realtek now works though it still post-resets to 160 even though I set 133 (anything lower crashes both), which sticks when using my Creative card: https://www.diffchecker.com/LCYzsRbR/

[kcat]

Can confirm 👌
Realtek now works though it still post-resets to 160 even though I set 133 (anything lower crashes both), which sticks when using my Creative card: https://www.diffchecker.com/LCYzsRbR/

Possibly the Realtek device can't go lower than 160 (~3.333ms), while the Creative card can go a bit lower. If initializing fails with AUDCLNT_E_BUFFER_SIZE_NOT_ALIGNED, it gets the next supported period size, which would be 160. Values less than 133 are probably triggering a different initialization error (AUDCLNT_E_BUFFER_SIZE_ERROR maybe?), which causes a failure and the delayed crash.

[ThreeDeeJay]

Replied in the main thread #682 (comment)