Add AEC support

[yujonglee]

AEC Model

• NKF-AEC
• DTLN-aec

We need inference code in Rust. There is Speexdsp binding, but not sure it will work well, especially in ASR context.

Mic/Speaker stream synchronization

Unless this issue is resolved:

1. Construct a StreamInstant from SpeakerStream. Since we already know mach_absolute_time, we can obtain a StreamInstant using this conversion.

2. StreamInstant is already available in MicStream(from Cpal).

3. Use duration_since to calculate the time delta and synchronize accordingly.

Not sure syncing is really needed. TDC is mentioned in NKF-AEC, but delay might be small enough for our case.

[ComputelessComputer]

Removing sounds from typing the keyboard would be suuuuuuper nice

[yujonglee]

In macOS, #904 can be solution.

[yujonglee]

We don't do sync right now. DTLN will be implemented #1007.