Fix circular DMA

[bjoernQ]

Fixes #1142

This fixes a subtle bug which was discovered in #1142

@thedevleon mind testing if this fixes your original issue?

[jessebraham]

Assuming this resolves the issue (pending results) this looks fine, so feel free to merge whenever.

[bjoernQ]

@jessebraham there are new clippy lints triggering as of today's nightly - and at least the thing about AesFlavour might be a false-positive. I wonder if we should limit clippy to the MSRV toolchain? We could get a bunch of new warnings/errors whenever we raise the MSRV then but with nightly we can get new (and especially unrelated) warnings out of a sudden with nightly - not sure what would be the best way to go ... I could probably mute those lints here but having totally unrelated changes in random PRs doesn't sound pleasant to me 🤔

[thedevleon]

@thedevleon mind testing if this fixes your original issue?

Yes, the embassy_i2s_sound sample now produces a noise-free sine wave without any additional modifications to the sample itself.