Allow compact notation to depend on style

[sffc]

Currently GetNotationSubPattern is a self-contained operation: https://tc39.es/ecma402/#sec-getnotationsubpattern

However, the notation subpattern can depend on the style. In ICU, this is already the case for currency, and it is likely useful for percent and unit styles as well.

Compact without currency:
https://unicode.org/cldr/charts/43/by_type/numbers.compact_decimal_formatting.html#76d709105676a7e5

Compact with currency:
https://unicode.org/cldr/charts/43/by_type/numbers.compact_decimal_formatting.html#122a473b23a455b1

By allowing this dependency, it could also give implementations more flexibility to use SI prefixes when style is "unit", instead of the default compact suffixes for that locale. CC @ljharb @ctcpip

[ctcpip]

not as confusing:

const numFormat = new Intl.NumberFormat('en-US', { style: 'unit', unit: 'byte', notation: 'compact' });

let num = 1;

for (let i = 0; i < 15; i++) {
  console.log(numFormat.format(num));
  num *= 10;
}

1 byte
10 byte
100 byte
1K byte
10K byte
100K byte
1M byte
10M byte
100M byte
1B byte
10B byte
100B byte
1T byte
10T byte
100T byte

more confusing:

const numFormat = new Intl.NumberFormat('en-US', { style: 'unit', unit: 'byte', notation: 'compact', unitDisplay: 'narrow' });

let num = 1;

for (let i = 0; i < 15; i++) {
  console.log(numFormat.format(num));
  num *= 10;
}

1B
10B
100B
1KB
10KB
100KB
1MB
10MB
100MB
1BB
10BB
100BB
1TB
10TB
100TB

[ljharb]

If the unit is "byte", why would anyone describe it in terms of millions/billions/trillions instead of mega/giga/tera?

[ctcpip]

see also: #730