Add dynamic/decode nullable_subfield decoder

[jhillyerd]

This is a spin-off of #809 and my related proposal.

This implements a nullable_subfield use-style decoder, which supports Nil values along the path; someting that optional_at does not.

Let me know if this seems reasonable and I will spend some time on docs and change-log. I went a little overboard on tests, happy to eliminate any that seem unnecessary.

[lpil]

Thank you but we do not accept unplanned PRs, any additions will need to be discussed beforehand. What's best here is still unclear to me, we will need to think through the options more.

[jhillyerd]

Re: discussions, I did that with my proposal linked above, but maybe it should have been a GH issue instead of a GH discussion.

Please consider adding a CONTRIBUTING.md with some guidance, given the lack of issue and PR templates in this repo, it's not easy to know where to start.