Fix out-of-order dimg grid associations

[y-guyon]

Fixes #2311.

A simpler alternative would have been to refuse to decode out-of-order dimg associations.

[y-guyon]

Note that the current implementation may not be as optimal as it could be for the following scenario:

• A part of an encoded grid image with out-of-order dimg associations is received at timeA
• libavif decodes all the in-order numO tiles among the numA available tiles
• The rest of the file is received at timeB
• libavif decodes all of the remaining tiles
• the image is fully decoded with a delay corresponding to min(timeB - timeA, timeRequiredToDecodeTheNumOTiles - timeRequiredToDecodeTheNumATiles) compared to eager decoding of every available tile at any time

This only concerns the unlikely out-of-order dimg associations use case and is not worth the fixing effort in my opinion.

[vigneshvg]

i will take a look at the rest of the PR in a while.

[wantehchang]

Yannis: Here is my first round of review. I saved some style/preference comments for the next round. I did not review the new avifdimgtest.cc file, but I reviewed the changes to read.c carefully.

Please check my first comment in read.c (marked with "IMPORTANT") carefully. If I am right, this pull request needs more work.

[wantehchang]

LGTM. Thanks. Please evaluate the optional suggestions and only accept the ones you agree with. There are two important changes that should be made.