Proposal: Redesign of `EntityEncoders` on using `Charset` in implicit scope

[danicheg]

This is a follow-up to a started discussion in https://github.com/http4s/http4s/pull/6528#discussion_r928091805. Since Charset isn't a typeclass, using it in implicit scope isn't that accurate. Though I haven't any ideas besides moving it to the explicit scope and creating an object with predefined UTF-8-based encoders (so we won't ask for Charset through an implicit scope within there). But yeah, this definitely will break a few existing implicit encoders. And sure, we can do such things only for 1.0. What are your thoughts about it? I can be super wrong here, so don't hesitate to point this out.