Consider relaxing alpha bit depth constraint

[y-guyon]

Section 4.1 says:

An AV1 Alpha Image Item (respectively an AV1 Alpha Image Sequence) shall be encoded with the same bit depth as the associated master AV1 Image Item (respectively AV1 Image Sequence).

This may have benefits for software or hardware implementations. However higher precision than 8 bits for alpha is rarely necessary.
I did not test whether the compression rate could be improved by sticking to 8-bit alpha samples along a 10+-bit color image, but restricting the range to useful values would not hurt on the file format design side.

[y-guyon]

Note that this was enforced in #74.

[leo-barnes]

Hmmm. That's a fair point. I have a vague memory that MIAF maybe says something about this. We should verify that. If it doesn't I don't really have any objections to relaxing it.

You're mainly arguing for allowing a smaller bit-depth for alpha. Do we want to keep the restriction that bit-depth of alpha shall be <= bit-depth of RGB?