refactor(toolchain): clean up toolchain hierarchy, pt. 2

[rami3l]

Part of #5092.

The main changes in this patch are:

• This patch has created a new trait ToolchainNameExt to accommodate the current Cfg::toolchain_path() and Toolchain::exists() functions. The rationale is:

  • These functions currently live in different impl blocks, however they are intended for different types of the toolchain name layer, which has been a source of confusion for me.
  • In addition, unfortunately, they currently only accept &ToolchainNameOrPath so values of other types would require a conversion to the former type first in order to to call these methods, which is not really necessary.

• DistributableToolchain has been renamed to ChannelToolchain to mirror ChannelToolchainName as previously discussed here.

  • The redundant ChannelToolchain::desc field has been removed, making ChannelToolchain a transparent newtype over Toolchain. This is a more conservative version of the original plan¹.

• To mirror the above type name change, mod toolchain::distributable has been renamed to mod toolchain::channels.

Footnotes

1. The original plan was to make the Toolchain type generic like so: struct Toolchain<'a, T: ToolchainNameExt = ToolchainNameOrPath>, so we can have type ChannelToolchain<'a> = Toolchain<'a, ChannelToolchainName>. Although this is completely feasible, after some experiments, I think this is making rustc quite unhappy because it is requiring explicit type annotations everywhere on Toolchain::new(), even when the original Toolchain::new() and the aliased ChannelToolchain::new() have completely incompatible output types. ↩

[rustbot]

This PR was rebased onto a different main commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.