Add formal wording for hash to indirect about value_types

[jbcoe]

@Ukilele I hope it was implicit that our answer to "I hope it's fine that I append my thoughts to this issue" is that your input is, has been and continues to be extremely welcome. You've a great eye for detail and are a pleasure to work with.

[Ukilele]

I took a look at the current implementation of hash at indirect.h#L288. I hope it's fine that I append my thoughts to this issue:

1. Do you intend to support hash for indirect with custom allocators? Right now std::hash is specialized only for indirect<T> (which is the same as indirect<T, std::allocator<T>>), but there is no support for arbitrary allocator types.
2. I took a look at how std::hash is specialized for std::optional: https://eel.is/c++draft/optional#hash. The wording also supports hashing std::optional<const T>. Do you intend to support hashing indirect<const T>, too? I think that before answering this question, the first question to answer is whether indirect<const T> (and polymorphic<const T>) should be supported.

[Twon]

Hi @Ukilele,

Thanks, good catch regarding std::hash, that's an oversight in the implementation. This should work with custom allocators, I've now added support for this: #56

Regarding whether we should support indirect<const T> and polymorphic<const T> that is a good question and should probably be mentioned either way in the paper. @jbcoe what are your thoughts here?