[breaking change] Make `utf8.encode()` / `Utf8Codec.encode()` return more precise `Uint8List` return type

[mkustermann]

Change Intent

Make Utf8Codec.encode() return a Uint8List instead of a List<int>.

Justification

This change will to allow compilers to better optimize code which is accessing the bytes. It also allows being explicit about the guarantees we make: We already guarantee to return a Uint8List but it's not visible in the return type. Some users then do explicit downcasts via utf8.encode(...) as Uint8List, others pass around List<int> which can result in inefficient code.

Impact

The impact will be more precision in static types and smaller code size / higher performance is some cases. It may break existing code if such code is relying on type inference to infer List<int> and would fail if it would now infer Uint8List.

Mitigation

Explicitly pass List<int> as type arguments in places where the automatically inferred Uint8List type is incorrect.

[lrhn]

This is a follow up on #36900 which intended to make that change, but ended up changing Utf8Encoder.convert instead.

(I do not remember the reason for not changing Utf8Codec.encode at the time, or whether it was just an oversight.)

[itsjustkevin]

@vsmenon @grouma and @Hixie for breaking change review.

[Hixie]

Very LGTM, been asking for this for years :-)

[kevmoo]

Ditto. Super useful change.