migrate remaining errors to new model

[asmello]

The EncodingError and ParseIndexError types are parsing errors with a
well defined 'subject', but they weren't previously made into Diagnostic
implementors, so they worked a bit inconsistently with other error types. This
PR makes it so those errors follow the same conventions and general structure
used by its kindred.

There is one related breaking change: InvalidCharacterError previously held
a copy of the subject, but in the new model this is a responsibility of its
enriched Report<InvalidCharacterError> type. Since some methods directly
depended on access to the subject, they had to be removed. I think this is
justifiable since the new error APIs are already breaking (we just should've
made this change in tandem with the other related breaks).

I don't have any other error-related breaks planned.

This PR also incidentally fixes an incorrect error type returned if the token ended with a ~ (it would be flagged as a slash error, not tilde). I updated tests to cover this.

[asmello]

TODO: update changelog

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 92.41706% with 16 lines in your changes missing coverage. Please review.

Project coverage is 95.8%. Comparing base (f3c23b9) to head (df39beb).

Files with missing lines	Patch %	Lines
src/index.rs	88.0%	13 Missing ⚠️
src/token.rs	96.9%	3 Missing ⚠️

Additional details and impacted files

Files with missing lines	Coverage Δ
src/assign.rs	97.6% <100.0%> (-0.1%)	⬇️
src/diagnostic.rs	77.9% <ø> (+3.4%)	⬆️
src/token.rs	98.7% <96.9%> (+1.0%)	⬆️
src/index.rs	91.2% <88.0%> (-1.0%)	⬇️

[chanced]

Damnit, I really regret pushing 0.7. Agreed that it should have been bundled.

We can cut a 0.8 in short order i think. I doubt many will have picked up 0.7 already and the breaks slotted for it won't be that bad.