fix: Update audio transcription example

[terryyylim]

I was exploring audio transcription on Baseten and attempted the tutorial, and found that these changes had to be made before I could get it working.

As referenced from StubBase documentation, we should be using self.predict_async and inputs parameter instead.

As referenced from DockerImage documentation, the base_image parameter should be either BasetenImage or CustomImage instead of a string.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}