clean up misc from retro-review of #101

[asmello]

I changed my mind on removing serde as a default feature. It is a heavy dependency to bring by default, but it's easy enough to disable the default features, and I do see the "common" use case for jsonptr involving serde_json. Also I guess Diagnose makes sense when you consider that the error it converts isn't yet a Report, even though it implements Diagnostic. Perhaps it's Diagnostic that requires reconsidering.

Changes I'm pushing forward:

• make diagnostic::Label::new private
• add docs do validate_bytes
• remove NoLeadingSlash struct
• update delete.rs docs
• deprecate ResolveError

Left for standalone PRs:

• Migrate index::ParseIndexError and token::EncodingError to the new error system.
• Seal Diagnose

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 95.9%. Comparing base (b302568) to head (f2677c8).

Additional details and impacted files

Files with missing lines	Coverage Δ
src/delete.rs	99.4% <ø> (ø)
src/diagnostic.rs	74.4% <ø> (ø)
src/pointer.rs	96.0% <ø> (+0.4%)	⬆️
src/resolve.rs	93.6% <ø> (ø)

[asmello]

This is technically breaking, as reported by the semver check, but nobody would be using the removed type because it's useless / not referenced anywhere.

[chanced]

make diagnostic::Label::new private

This one needs to be public in order for users to implement Diagnostic properly. Users can bring their own errors for Assign and Resolve but they have to implement Diagnostic.

remove NoLeadingSlash struct

Ugh, thank you. Sorry I left that in.

[asmello]

This one needs to be public in order for users to implement Diagnostic properly. Users can bring their own errors for Assign and Resolve but they have to implement Diagnostic.

Ah, interesting, but do we expect users to implement Diagnostic themselves?

[chanced]

I dont necessarily expect it - I doubt anyone is implementing Assign and Resolve for custom languages or data structures - but it is possible and in an effort to facilitate that, the traits were designed so that implementations bring their own error types for resolve::Resolve and assign::Assign.

[asmello]

You know what, you're completely right. I lost track of what those traits were for, it does make sense that people can use jsonptr with their own types. Although, now that I think about it, it also only makes sense for JSON-like data structures.

I'll update this PR and remove that change. Thanks!