Support ChainRulesCore v1 and ChainRulesTestUtils v1

[sethaxen]

Supersedes #25, #26, and #28

[sethaxen]

The tests currently fail because ChainRulesTestUtils.test_approx doesn't know what to do with the QDHT type (see example). @mzgubic, what's the right way in ChainRulesTestUtils to use test_frule and test_rrule on constructors?

[mzgubic]

The usage looks right to me.

The error we are seeing is fixed by defining Base.:(==) for QDHT.

The next error comes from Tangent{QDHT} and QDHT not having the same length when to_vecd. This could be fixed by JuliaDiff/FiniteDifferences.jl#188 (but doing that might mess up some other ChainRules tests - I haven't checked yet.)

Doing both fixes still fails the tests, namely the finite differencing thinks the R is actually differentiable. I don't know anything about Hankel so not sure if that makes sense? In case the derivative exists but hasn't been worked out yet @not_implemented macro can be used.

My suggestion would be to use the existing tests and open an issue to use test_f/rrule