rustc_session: Add a structure for keeping both explicit and default sysroots

[petrochenkov]

Also avoid creating and cloning sysroot unnecessarily.

Implements the suggestion from #142089 (comment).
r? @bjorn3

[rustbot]

rustc_error_messages was changed

cc @davidtwco, @compiler-errors, @TaKO8Ki

Some changes occurred in compiler/rustc_codegen_ssa

cc @WaffleLapkin

These commits modify the Cargo.lock file. Unintentional changes to Cargo.lock can be introduced when switching branches and rebasing PRs.

If this was unintentional then you should revert the changes before this PR is merged.
Otherwise, you can ignore this comment.

[bjorn3]

Maybe return &[PathBuf] instead? Then fluent_bundle won't need to be made generic.

[petrochenkov]

How? Can't do that without changing the storage in Sysroot to something like SmallVec<[PathBuf; 2]>.

fluent_bundle could also be made non-generic by collecting the iterator at its call site.

[bjorn3]

Right

[bjorn3]

I've been meaning to split --sysroot into a host sysroot (for codegen backends and executables) and target sysroot (for the standard library and self-contained libraries) option and then only access the default sysroot when the respective option is not set. The current fallback behavior is necessary because someone may pass --sysroot to override the standard library without copying all host executables and libraries into the passed sysroot. If the host and target sysroot can be configured separately, fallback for missing artifacts shouldn't be necessary anymore. Until that is implemented, I do think this PR is an improvement though.

r=me with the all_paths return type change.