Incremental decoding implemented

[birktj]

This is an implementation of incremental decoding (decoding one strip at a time). This PR is the first step towards image-rs/image#858. I am reasonably confident that the implementation is correct in the same cases as the last implementation. I have tested it on the tests from #25 and they are correct.

The public api has changed in a small way with two new public methods on Decoder, namely Decoder::strip_count and Decoder::read_strip. If there are any objections to the api changes or the implementation I would be happy to sort them out.

[birktj]

@fintelia (or anyone else) I wonder if you could give me any feedback on this PR so that I can continue with image-rs/image#858?

[fintelia]

I'm not really that familiar with the TIFF decoding logic, but assuming that tests pass and so forth the code looks reasonable. My one concern is that having read_strip allocate and return a Vector might cause you to have to make extra copies when implementing Read.

If you think this is ready, I can go ahead and merge this PR and we can address any rough edges in a future patch.

[birktj]

@fintelia I tried to keep it compatible with the current DecodingResult. I believe this could be done better, but that would require some larger api changes which is something I am looking into for the longer run.

I guess if I marked DecodingBuffer and read_strip_to_buffer as public I could fix the extra copy issue and simply copy some of the read_strip logic into the eventual Reader. I will fix that now and then we should then have what we need to continue with an TiffReader in the image crate for now.

[birktj]

@fintelia friendly ping. Would it be possible to merge this or do you have any other concerns.