Upgrade syn to 2.0

[jplatte]

Didn't convert the huge macro_rules! macro and some non-trivial NestedMeta usage. I'd rather just convert it to parsing in one step like I did for match_attribute_with_str_list_value, but (a) I expected this to be a short contribution and have run out of time for now and (b) I'm not sure this is even what you want, so let's discuss how to proceed from this draft PR.

Resolved #401.

[zeenix]

@bilelmoussaoui any chance you can take over?

[zeenix]

I'm not sure this is even what you want,

Can you please elaborate on this?

[jplatte]

I'm talking about this change. It impacts error messages, e.g. supplying attribute_argument = true for an attribute_argument that expects a string value would previously result in a compilation error saying "invalid literal type for attribute_argument attribute", and now will likely say something like "syntax error, expected "" (pointing to true). Are you okay with error reporting changing to syn's parse errors, from the custom errors based on a parse result?

[zeenix]

Thanks for explaining. 👍

Are you okay with error reporting changing to syn's parse errors, from the custom errors based on a parse result?

Yeah. The new error seems more descriptive even. My preference would always bee what's most helpful to the user.

[jplatte]

This was the hardest piece to find out. Apparently syn 1.x just accepted ImplItemMethods without a body, but syn 2.x does not. The error message when re-emitting this ImplItem::Verbatim was plenty confusing (it was "expected impl" on the first token of the item).

[zeenix]

Closing in favour of #798.