After merging main (which brought in PR #352's "Doc comment intra-links"
guide and PR #351's "Documentation comments / private item refs" guide),
re-audit the bin-linking docs against the new sections and convert two
remaining bare-prose item mentions:

* `crates/cmd-shim/src/capabilities.rs:23` — the `FsReadHead` trait doc
  said "via `read_head_filled`" with the helper's name as bare prose.
  Convert to [`crate::read_head_filled`].
* `crates/package-manager/src/link_bins.rs:69` — the `LinkVirtualStoreBins`
  struct doc named `create_symlink_layout` as bare prose. The function
  shares its name with the module that hosts it, so the link needs to
  disambiguate; use `[create_symlink_layout](crate::create_symlink_layout())`
  to point at the function form.

The `Command` mention in `bin_resolver.rs:9` deliberately stays as bare
backticks: it refers to *upstream pnpm's* `Command` class (in
`@pnpm/bins.resolver`), not the local struct, so the bare-prose form is
the accurate reading.

…links

Apply the convention added by #352: when a doc comment
mentions an identifier that is rustdoc-linkable from the current scope,
write it as an intra-doc link rather than bare backticks. Two callouts
in `crates/modules-yaml/src/lib.rs` qualify (`serde_json::to_string_pretty`
and `serde_json::Value`); both are public re-exports of in-tree
dependencies, so rustdoc can resolve them and a renamed/removed item
will surface as a `cargo doc` warning instead of stale prose.

`cargo doc --no-deps` runs clean with `-D
rustdoc::broken_intra_doc_links -D rustdoc::private_intra_doc_links`,
confirming no `///` doc references a more-private item per the
guidance in #351.

Follow the new style rules from #352 (use rustdoc intra-doc links
for in-scope identifiers in doc comments) across the auth-related
files added by this PR.

Notable fixes:
* `crates/npmrc/src/env_replace.rs` had a non-resolving
  `[\`EnvReplaceError::Missing\`]` reference (`EnvReplaceError` is a
  struct, not an enum). Drop the bogus `::Missing` segment and link
  the type itself.
* Test docstrings across `auth.rs`, `npmrc_auth.rs`, and `lib.rs`
  now use intra-doc links for the in-scope items they reference
  (`nerf_dart`, `ParsedUrl::parse`, `creds_to_header`,
  `split_creds_key`, `apply_creds_field`, `NpmrcAuth::*`,
  `Npmrc::current`, `EnvVar::var`, `env_replace`, etc.).
* Registry test docstrings link `Package::fetch_from_registry` and
  `PackageVersion::fetch_from_registry` rather than naming them in
  bare backticks.

PR 351 ("don't reference private items from public docs") audited
in the same pass; the new pub items in this PR only reference other
pub items in their docs, so no changes were needed for that rule.
The integration test in `crates/cli/tests/install.rs` keeps bare
backticks because rustdoc does not process integration-test files
and the imported scope there does not include the referenced
`pacquet_npmrc` items.

After merging main (which brought in PR #352's "Doc comment intra-links"
guide and PR #351's "Documentation comments / private item refs" guide),
re-audit the bin-linking docs against the new sections and convert two
remaining bare-prose item mentions:

* `crates/cmd-shim/src/capabilities.rs:23` — the `FsReadHead` trait doc
  said "via `read_head_filled`" with the helper's name as bare prose.
  Convert to [`crate::read_head_filled`].
* `crates/package-manager/src/link_bins.rs:69` — the `LinkVirtualStoreBins`
  struct doc named `create_symlink_layout` as bare prose. The function
  shares its name with the module that hosts it, so the link needs to
  disambiguate; use `[create_symlink_layout](crate::create_symlink_layout())`
  to point at the function form.

The `Command` mention in `bin_resolver.rs:9` deliberately stays as bare
backticks: it refers to *upstream pnpm's* `Command` class (in
`@pnpm/bins.resolver`), not the local struct, so the bare-prose form is
the accurate reading.