rust-lang/rust#121621

    warning: non-local `impl` definition, they should be avoided as they go against expectation
        --> tests/test.rs:2338:5
         |
    2338 | /     impl<'de> Deserialize<'de> for &'de RawMapKey {
    2339 | |         fn deserialize<D>(deserializer: D) -> Result<Self, D::Error>
    2340 | |         where
    2341 | |             D: serde::Deserializer<'de>,
    ...    |
    2345 | |         }
    2346 | |     }
         | |_____^
         |
         = help: move this `impl` block outside the of the current function `test_raw_value_in_map_key`
         = note: an `impl` definition is non-local if it is nested inside an item and neither the type nor the trait are at the same nesting level as the `impl` block
         = note: one exception to the rule are anon-const (`const _: () = { ... }`) at top-level module and anon-const at the same nesting as the trait or type
         = note: this lint may become deny-by-default in the edition 2024 and higher, see the tracking issue <rust-lang/rust#120363>
         = note: `#[warn(non_local_definitions)]` on by default

Remove conditional on repr(transparent)

New in nightly-2024-03-24 from rust-lang/rust#119552.

    warning: field `x` is never read
      --> tests/regression/issue795.rs:11:15
       |
    11 |     Variant { x: u8 },
       |     -------   ^
       |     |
       |     field in this variant
       |
       = note: `#[warn(dead_code)]` on by default

    warning: field `i` is never read
      --> tests/regression/issue845.rs:63:9
       |
    61 | pub struct Struct {
       |            ------ field in this struct
    62 |     #[serde(deserialize_with = "deserialize_integer_or_string")]
    63 |     pub i: i64,
       |         ^
       |
       = note: `Struct` has a derived impl for the trait `Debug`, but this is intentionally ignored during dead code analysis

This was leading to rustdocs markdown parser to interpret slices like &[u8] as
& and a link to u8

Debugging a recent cargo-outdated bug, it would have been nice not to
wonder whether a rustc version change in GitHub's runner image was a
contributing factor.

Fix missing backtick in doc comments letting markdown think &[u8] is a link to u8