Add object-literal plugin types and deprecate class-based plugins

[knowledgecode]

Summary

• Add FormatterPluginObject / ParserPluginObject, which let a plugin be written as a plain object literal (the v4 counterpart of 3.x date.extend(...)) and statically reject keys that collide with built-in tokens.
• Rewrite the built-in formatter/parser and the nine bundled plugins as object literals. The reserved token set is an explicit union (ReservedFormatterToken / ReservedParserToken), and satisfies Record<Reserved…, …> on the built-in objects keeps it exactly in sync with their keys. The bundled plugins use satisfies rather than a type annotation, so each token keeps its own function type instead of widening to a string-indexed record.
• Mark FormatterPlugin / ParserPlugin as @deprecated. The plugins option now accepts (FormatterPluginObject | FormatterPlugin)[] (likewise for the parser), so existing class-based plugins keep compiling until the classes are removed in the next major version.
• Docs: add a "Writing Your Own Plugin" section and a 3.x to 4.x migration example, update the plugins option type in the format/parse API docs, and clarify that only annotated object literals are checked against built-in tokens, what a parser token can and cannot do, and that the class-based types are deprecated.
• Chore: bump the version to 4.6.0 and update devDependencies (vitest / @vitest/coverage-v8 5.x, astro 7.3, @astrojs/starlight 0.42).
• Docs build: fix the two warnings that appeared after the Astro/Starlight update (empty i18n collection, missing 404 entry). docs/404.md is added and disable404Route is set, so the 404 page is generated from a real entry.
• Tests: enable fsModuleCache for faster local reruns and set isolate: true explicitly. The tests change process.env.TZ, which would leak across files if workers were shared.

Test plan

• npm run ts
• npm run lint
• npm run test (72 files / 2310 tests)
• npm run build
• npm run docs:build (no warnings)
• Type-check the built dist/*.d.ts from a consumer's perspective: object literals, class instances, and bundled plugins (with their per-token types) are accepted; a key redefining a built-in token and a non-function token are rejected

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.78%. Comparing base (f2495b0) to head (5a2069b).

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #139   +/-   ##
=======================================
  Coverage   98.78%   98.78%           
=======================================
  Files          77       77           
  Lines        1233     1235    +2     
  Branches      357      357           
=======================================
+ Hits         1218     1220    +2     
  Partials       15       15           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.