Editorial: introduce CanonicalizeKeyedCollectionKey

[bakkot]

Per #3306 (comment).

[michaelficarra]

Oooh, nice!

[gibson042]

Thanks! This could also be used by the related algorithms that currently rely upon SameValueZero.

           1. For each element _e_ of …, do
-            1. If _e_ is not ~empty~ and SameValueZero(_e_, _value_) is *true*, …
+            1. If _e_ is not ~empty~ and CanonicalizeKeyedCollectionKey(_value_) is _e_, …

or

           1. For each element _e_ of …, do
-            1. If _e_ is not ~empty~ and SameValueZero(_e_, _value_) is *true*, …
+            1. If CanonicalizeKeyedCollectionKey(_value_) is _e_, …

[ljharb]

I like this AO, but that specific suggestion seems like it'd reduce clarity to me - i'd think it'd be better to retain separate "canonicalize" and "comparison" AOs.

[michaelficarra]

@gibson042 I certainly wouldn't want to replace all uses of SameValueZero, but I'd be fine doing that for the ones in Map/Set.

[bakkot]

This could also be used by the related algorithms that currently rely upon SameValueZero.

Done, though I've written it as canonicalizing before the loop rather than in every step of the loop.

It's technically slightly redundant to write If _e_ is not ~empty~ and _e_ is _value_ because _value_ is never ~empty~, but I've chosen to keep the first half to emphasize that holes are being skipped here.