Add -typederrors flag for typed enum conversion errors

[samiam2013]

Summary

Implements the typed errors feature discussed in #111 using the errors.Join() approach. This provides type-safe error checking while maintaining full backwards compatibility.

Key Features

• New -typederrors flag: Optional flag that enables typed error generation (defaults to false)
• Custom error type: enumerrs.ErrInvalidValue sentinel error for type-safe checking
• errors.Join() approach: Combines typed error with descriptive messages
• Full backwards compatibility: Existing code continues working unchanged

Implementation Details

• With -typederrors: Returns joined typed error errors.Join(enumerrs.ErrInvalidValue, fmt.Errorf("..."))
• Without flag: Returns standard fmt.Errorf(...) (unchanged behavior)
• All marshaling methods supported: JSON, YAML, Text unmarshaling all use typed errors when enabled
• Test coverage: All existing tests pass, new functionality verified

Test Plan

• Verify typed error detection with errors.Is()
• Verify backwards compatibility without flag
• Verify descriptive error messages preserved
• Verify all existing tests continue to pass
• Verify JSON/YAML/Text unmarshaling uses typed errors when enabled
• New Golden Test
• New github CI test flows for e2e & golden testing

🤖 Generated with help from Claude Code

[samiam2013]

Thank you for your review @bincyber I plan on rebasing and adding you as a contributor for your efforts here unless you object. I think in addition to the change to a single shared definition of this error it should be named more idiomatically, if I'm remembering correctly that means ErrInvalidValue with the err first since it's a value and primarily an error.
I appreciate your enthusiasm for this change and I will be trying to find the time to fit this in but you should know my best friend IRL has some things I need to help with and it will be a juggling act to fit this into the rotation

[bincyber]

Thank you for your review @bincyber I plan on rebasing and adding you as a contributor for your efforts here unless you object. I think in addition to the change to a single shared definition of this error it should be named more idiomatically, if I'm remembering correctly that means ErrInvalidValue with the err first since it's a value and primarily an error. I appreciate your enthusiasm for this change and I will be trying to find the time to fit this in but you should know my best friend IRL has some things I need to help with and it will be a juggling act to fit this into the rotation

Sounds great

[samiam2013]

I went to add this error globally and found it can't be at the top of the repo because that's where the tool code lives and we would have to make enumer itself a depdendency of the generated code. I was able to sort an implementation of the change using strconv.ErrRange.

One alternative for having the error in scope without adding a dependency is to de-duplicate: we could write just a separate file with the error type (typeS in the future) and check if it's there before writing again from another type being generated.

This has downsides like the errors could be duplicated across a codebase in different paths and even just different levels of hierarchical data structures. It would also be a more significant refactor.

I think the path of a new package space inside enumer, github.com/dmarkham/enumer/errs is the least-friction path ahead and because this addition of a generated non-standard-library dependency is behind a flag that's newly added I don't think it represents the concern it felt like when I uncovered that requirement.

I also need to remember to update the README with this new flag

[bincyber]

🚢

[samiam2013]

One last thing: like #111 I think this deserves a "golden" test for the generated return error

[samiam2013]

because this is an addition but is backwards compatible, I think the new tag should be version 1.6.0

[samiam2013]

I keep thinking about how to test this further and because the referenced error is in this change, I can't do too much. I know it's working because of the testing, and I would have noticed the other .golden test files getting broken if it broke backwards compatibility. I'm just going to merge it.