Add StreamReadBits() and StreamWriteBits()

[y-guyon]

Simplify calling sites that were doing manual bit manipulation.
Also add VarInt versions.

#1432 will be simplified when this PR is merged.

[maryla-uc]

Nice! I'm wondering if the other, whole byte writing/reading functions should check that numUsedBitsInPartialByte is 0? to make sure that padding bits were properly added? otherwise if I understand correctly, the next bit read/write will keep acting on a previous byte which seems like a weird api? but maybe that's intentional or maybe I didn't read carefully enough

[y-guyon]

Nice! I'm wondering if the other, whole byte writing/reading functions should check that numUsedBitsInPartialByte is 0?

avifROStreamSetOffset() already had assert(stream->numUsedBitsInPartialByte == 0);. I added the assertion to other functions. offsetOfPartialByte made no sense anymore (as described in #1432 (comment)) so I removed it. I fixed avifstreamtest too because it tested that pattern.

otherwise if I understand correctly, the next bit read/write will keep acting on a previous byte which seems like a weird api

This is the behavior of libwebp2 for what it is worth.

but maybe that's intentional or maybe I didn't read carefully enough

Yes it was intentional. I will add it back one day if needed.

[maryla-uc]

Thanks for the details. I didn't realize webp2 did this. I think it can work but it's not really in the spirit of isobmff boxes. You could have bits leaking into other boxes which seems weird. Although in practice we're following what the specs say anyway, and they use padding bits. I think it's less error prone to check that padding was indeed properly added (like you did).