Implement calendrical calculations (meta-issue)

[sffc]

We decided in #355 to take pre-resolved calendar data when formatting. However, Temporal and other downstream clients will want to hand ICU4X an ISO date and have us convert that to a calendar-specific date. Our team is in a particularly good spot to implement this logic.

Outline:

• Calendrical calculations should be a separate crate from icu_datetime
• Ultra-fast conversion from ISO date to calendar date and back to align with ECMAScript Temporal -- Julian day or similar epoch day could be used internally, but ISO conversion is the performance-critical code path
• Standalone date type for each calendar system with getters for all date fields (e.g., GregoryDate, ChineseDate, BuddhistDate) -- polymorphism is NOT required since the ISO date should be used for interchange
• Support for arithmetic of years and months, as well as anything else required to implement the Temporal proposal

I made this meta-issue because I believe several smaller issues can be spun off to implement each individual calendar system. I believe each calendar system will need its own code path, although some may be able to share bits of code where possible. Most calendar systems can be implemented algorithmically, although come need data (e.g., the religious Islamic calendar); we should probably hook up DataProvider for this data.

Related: #493, which is about formatting data after we pass through the date input layer.

---

Sub-issues:

• Add initial calendars crate #827
• Implement formatting for a non-gregorian calendar #1115
• Implement the Japanese calendar #1116
• Write common framework for calendar arithmetic #1470
  • Implement the Coptic calendar #1119
  • Implement the Indian national calendar #1117
• Implement the Buddhist calendar #1118

[justingrant]

One suggestion for whoever ends up implementing this feature: there's an placeholder implementation of calendar calculations in the existing Temporal polyfill. The implementation itself IMHO doesn't have much value (it's inefficient and poorly written-- I know because I wrote it!), but it'd probably be useful to review the existing impl before building the real one. Specifically:

• Existing test cases are insufficient for a comprehensive test suite, but may be useful as a starting point
• The metadata and patterns to classify calendars (and how they're similar) used in the Temporal placeholder code are different from ICU4*'s existing logic. There might be some useful info in how the implementation defines eras or month codes, for example.
• There are bugs in ICU logic that are called out in the current placeholder implementation and tests; we should make sure that those bugs are fixed in the real version.

[Manishearth]

Design doc for common API at https://docs.google.com/document/d/1SrI7l13bi8trFh4882_n1rHebgszzCc0M2mpMuslIMA/edit?usp=sharing

[Manishearth]

Initial work at #827

[Manishearth]

#1018 is a subpart of this

[catull]

Design doc for common API at https://docs.google.com/document/d/1VVydmguGNCTEjFVhYVl9d0EBH34PfIETh3xCVgepzwE/edit?resourcekey=0-NcWu-l38IfZ4ytz8VBZ3ZQ#

This link is outdated, please edit your comment above, @Manishearth.
Thanks.