Avoid returning to ReadChunkData state

[kornelski]

The chunk size is known ahead of time, and it's always fully read, so there's no need to have a separate state for growing it incrementally.

With the simplified state, it would also be possible to move parsing of chunks after the crc check.

[fintelia]

This might be better to defer until after we've gotten 0.18.0 out.

It has a few consequences that are a bit hard to think through: it allocates space for entire chunks ahead of time rather than waiting to see whether we hit an EOF first and it mem::take's vectors in certain places without a shrink_to_fit first, so their capacity could be far larger than their current length.