HW/DVDInterface: Avoid heap allocation in DTK callback.

[AdmiralCurtiss]

This should be equivalent, and 672 bytes is small enough that having it on the stack should be fine.

[delroth]

One small comment, because I don't think there's a good reason to leave the risk of overflowing a stack buffer here.

[AdmiralCurtiss]

tbqh this still seems a bit shaky, this definitely breaks if m_pending_samples ever happens to be not divisible by SAMPLES_PER_BLOCK or audio_data.size() is ever not divisible by ONE_BLOCK_SIZE. This is also the case in master, but maybe worth refactoring anyway...

[AdmiralCurtiss]

I've rewritten the math to work in terms of DTK data blocks instead of sample counts, this should be safer in case we ever get a weird value somehow. Please verify.

[delroth]

I'm not convinced moving to blocks made things any simpler... but whatever, either way is fine.