fix!: refactor v7 internal state and options logic, fixes #764

[broofa]

[#776, reopened as a new PR. No idea why it's including so many commits that are already on main]

Putting this PR up as an alternative to #768 that addresses #764 and a few other concerns.

• Written in Typescript (required now that feat!: Port to TypeScript, closes #762 #763 has landed)
• Decouple behavior of v7()-with-options from v7()-without-options.
• Encapsulate internal state handling and refactor to be more idiomatic
• Rigorous unit tests for internal state transitions
• Use 32-bit seq field rather than 31.

@robinpokorny My apologies for closing #768. I wanted to merge it, but the switch to typescript and these other issues just made it easier to address in a new PR. I'll do my best to credit your work here.

... and if you could help review this, I'd appreciate it. I feel pretty good about this refactor but there's enough code churn that an extra set of eyes would be very helpful.

Where #764 is concerned, the heart of this PR is in this code, where the behavior very deliberately branches based on whether options are present.

Note

• 'Marking this as a breaking change because of the changes to default values for options#seq and options#msecs.
• The v1() code would probably benefit from a similar refactoring, if for nothing else than to make the APIs consistent. I'll put up a PR for that shortly.

[robinpokorny]

I really like it now! I think the trade offs are well selected. Good job 🎉

Maybe, in README we should write about the monotonicity and that providing any options means no guarantees of it. WDYT?

[broofa]

@robinpokorny

Maybe, in README we should write about the monotonicity and that providing any options means no guarantees of it. WDYT?

I'll update the README once this PR and #780 land.