Hardware
I'm not 100% on this, I only write apps, rather than deep phone OS stuff, but my understanding is that the task of handling a phone call is done by the phone's hardware, rather than the OS.
Something like:
Call comes in via radio chip, which alerts the OS which prompts the user. The user accepts and the OS informs the hardware. The hardware then enables the mic & speaker until the call is terminated or the OS tells the hardware to terminate.
In which case, even though Android does support the codec, the phone's hardware also needs to support it. This way you don't get CPU lag during a call if you're hammering your phone (think old WinMo phones that could barely show the Start Menu without lagging.)
The hardware may provide methods for the OS to pick up the audio stream (for call recording or what not,) but I reckon that varies by handset/manufacturer.
That is my understanding anyway, but like I say, I haven't read up on it in any great detail.