Editorial: Use named records for DateTimeFormat format records

[anba]

This PR aims to restructure Intl.DateTimeFormat Constructor - Internal Slots to use named Record types instead of ad-hoc records. In the process of this clean-up, some editorial bugs were found and fixed, see the commit messages for details.

In addition to providing a more type-safe approach when compared to using ad-hoc records, these changes should also help with the Temporal integration, see https://tc39.es/proposal-temporal/#sec-temporal-intl.

---

The PR applies on top of #822.

[ben-allen]

Whew! Reviewed commits one-by-one; most of the comments are about things fixed in later commits. I've left those comments in place, but flagged them as fixed later.

Thank you again for all your work on this -- it's awe-inspiring.

[anba]

Fixed all review comments and rebased the commits to apply cleanly on tip.

[anba]

Rebased again.

Do we still wait for another review or can we land these changes?

[sffc]

@gibson042 If this looks good please merge.

[gibson042]

This is great stuff, but see my comments about surfacing "pattern" in the new names.

[gibson042]

I have a goal of replacing each [[rangePatterns]] and [[rangePatterns12]] unit-keyed Record with a List of Records with fields like { [[BiggestDivergentUnit]]: string, [[RangePattern]]: DateTimeFormat Range Pattern Record } Records (or possibly just a list of range pattern Records that include that first field, saving a level of depth), and it might be appropriate to include that change in this PR rather than introduce new bibliographic terms that will just need to be removed. Thoughts?

The net effect should be to reduce opacity of field semantics, replacing e.g.

[[rangePatterns]]: {
  [[Hour]]: {
    [[hour]]: "numeric",
    [[minute]]: "numeric",
    [[PatternParts]]: […]
  },
  [[Minute]]: {
    [[hour]]: "numeric",
    [[minute]]: "numeric",
    [[PatternParts]]: […]
  },
  [[Default]]: {
    [[year]]: "2-digit",
    [[month]]: "numeric",
    …,
    [[PatternParts]]: […]
}

with

[[rangePatterns]]: [
  { [[BiggestDivergentUnit]]: "hour",
    [[hour]]: "numeric",
    [[minute]]: "numeric",
    [[PatternParts]]: […]
  },
  { [[BiggestDivergentUnit]]: "minute",
    [[hour]]: "numeric",
    [[minute]]: "numeric",
    [[PatternParts]]: […]
  },
  { [[BiggestDivergentUnit]]: ~empty~,
    [[year]]: "2-digit",
    [[month]]: "numeric",
    …,
    [[PatternParts]]: […]
  }
]

[gibson042]

I'm also thinking about flattening these from [[DateTimeRangeFormat]].[[<dateStyle>]].[[<timeStyle>]] to [[DateTimeRangeFormat]].[[<dateAndTimeStyle>]] with e.g. Let _dateAndTimeStyle_ be the string-concatenation of _dateStyle_, *"-"*, and _timeStyle_., but haven't yet decided if that's worthwhile.

[gibson042]

😍

[ben-allen]

@anba I've rebased this onto the current main on my own fork: https://github.com/ben-allen/ecma402/tree/anba-826-working-branch. If you pull it in, I can merge.

[ben-allen]

@anba Rebased again! It's still at https://github.com/ben-allen/ecma402/tree/anba-826-working-branch

(also, apologies for letting this one sit so long)

[gibson042]

Suggested change

	1. Let _dateTimeFormat_ be ? OrdinaryCreateFromConstructor(_newTarget_, *"%DateTimeFormat.prototype%"*, « [[InitializedDateTimeFormat]], [[Locale]], [[Calendar]], [[NumberingSystem]], [[TimeZone]], [[HourCycle]], [[DateStyle]], [[TimeStyle]], [[DateTimeFormat]], [[BoundFormat]] »).
	1. Let _dateTimeFormat_ be ? OrdinaryCreateFromConstructor(_newTarget_, *"%DateTimeFormat.prototype%"*, « [[InitializedDateTimeFormat]], [[Locale]], [[Calendar]], [[NumberingSystem]], [[TimeZone]], [[HourCycle]], [[DateStyle]], [[TimeStyle]], [[Patterns]], [[BoundFormat]] »).

[gibson042]

This id seems superfluous; linking to the dedicated section should be fine.

Suggested change

	Each <dfn id="datetimeformat-patterns-record" variants="DateTimeFormat Patterns Records">DateTimeFormat Patterns Record</dfn> has the fields defined in <emu-xref href="#table-datetimeformat-patterns-record"></emu-xref>.
	Each <dfn variants="DateTimeFormat Patterns Records">DateTimeFormat Patterns Record</dfn> has the fields defined in <emu-xref href="#table-datetimeformat-patterns-record"></emu-xref>.

[gibson042]

Suggested change

	<td><emu-not-ref>Contains</emu-not-ref> for each of the date and time format component fields of the record a substring starting with *"{"*, followed by the name of the field, followed by *"}"*. If the record has a [[year]] field, the string may contain the substrings *"{yearName}"* and *"{relatedYear}"*.</td>
	<td><emu-not-ref>Contains</emu-not-ref> for each of the date and time format component fields of the Record a <emu-not-ref>substring</emu-not-ref> starting with *"{"*, followed by the name of the field, followed by *"}"*. If the Record has a [[year]] field, the string may contain the substrings *"{yearName}"* and *"{relatedYear}"*.</td>

[gibson042]

Suggested change

	<td>Each record represents a part of the range pattern.</td>
	<td>Each Record represents a part of the range pattern.</td>

[gibson042]

Suggested change

	<tr>
	<td>[[full]]</td>
	<td>a DateTime Format Record</td>
	<td>Format record for the *"full"* style.</td>
	</tr>
	<tr>
	<td>[[long]]</td>
	<td>a DateTime Format Record</td>
	<td>Format record for the *"long"* style.</td>
	</tr>
	<tr>
	<td>[[medium]]</td>
	<td>a DateTime Format Record</td>
	<td>Format record for the *"medium"* style.</td>
	</tr>
	<tr>
	<td>[[short]]</td>
	<td>a DateTime Format Record</td>
	<td>Format record for the *"short"* style.</td>
	</tr>
	<tr>
	<td>[[full]]</td>
	<td>a DateTime Format Record</td>
	<td>Format Record for the *"full"* style.</td>
	</tr>
	<tr>
	<td>[[long]]</td>
	<td>a DateTime Format Record</td>
	<td>Format Record for the *"long"* style.</td>
	</tr>
	<tr>
	<td>[[medium]]</td>
	<td>a DateTime Format Record</td>
	<td>Format Record for the *"medium"* style.</td>
	</tr>
	<tr>
	<td>[[short]]</td>
	<td>a DateTime Format Record</td>
	<td>Format Record for the *"short"* style.</td>
	</tr>

[ben-allen]

Rebased again! Most of the comments above are addressed by #887 -- I think the only one that isn't is the use of id in the DateTimeFormat Patterns Record dfn

@anba @gibson042

[ben-allen]

I've merged the rebased version of this PR and #827 via the new #893.