Grammar check: update rustc for forced keywords

[DanielEScherzer]

Fixes #2368
Cross ref rust-lang/rust#161775
Dispatched workflow confirmed the fix: https://github.com/DanielEScherzer/reference/actions/runs/36614349553/job/109563438649

[DanielEScherzer]

To be clear, this fixes the compilation, the changed grammar means that the actual grammar check fails

------------------------------------------------------------
error: reference failed, rustc_parse passed
test: rust/tests/ui/parser/forced-keywords/non-keyword-neg.rs
reference error: invalid token RESERVED_TOKEN at `k#not_a_keyword()
    //~^ ERROR expected expression, found `k#not_a_keyword`
}
fn k#trying_to_pass_…`
8:     k#not_a_keyword()
       ━
------------------------------------------------------------
error: reference failed, rustc_parse passed
test: rust/tests/ui/parser/forced-keywords/non-keyword-pos.rs
reference error: invalid token RESERVED_TOKEN at `k#not_a_keyword); // OK
#[cfg(false)]
invoke!(k#arbitrary_string_of_chars);
fn main() {}
`
9: discard!(k#not_a_keyword); // OK
            ━
------------------------------------------------------------
error: reference failed, rustc_parse passed
test: rust/tests/ui/parser/forced-keywords/decl-macro-matching.rs
reference error: invalid token RESERVED_TOKEN at `k#type) => {};
    (Normal type) => {};
}
accept!(ForcedKeyword type); //~ ERROR no rules expected …`
8:     (ForcedKeyword k#type) => {};
                      ━
------------------------------------------------------------
error: reference failed, rustc_parse passed
test: rust/tests/ui/parser/forced-keywords/basic.rs
reference error: invalid token RESERVED_TOKEN at `k#mod module {
    k#pub(k#in k#super) k#static _DATA: i32 = 0i8 k#as k#_;
}
k#use ::std::process::…`
5: k#mod module {
   ━
------------------------------------------------------------
error: reference failed, rustc_parse passed
test: rust/tests/ui/parser/forced-keywords/feature-gate-forced-keywords.rs
reference error: invalid token RESERVED_TOKEN at `k#fn main() {} //~ ERROR forced keywords are experimental
#[cfg(false)]
k#fn start() {} //~ ERROR f…`
3: k#fn main() {} //~ ERROR forced keywords are experimental
   ━
------------------------------------------------------------
error: reference failed, rustc_parse passed
test: rust/tests/ui/parser/forced-keywords/pre-2021-edition-fail.rs
reference error: invalid token RESERVED_TOKEN at `k#fn main() {} //~ ERROR expected one of `!` or `::`, found `#`
`
3: k#fn main() {} //~ ERROR expected one of `!` or `::`, found `#`
   ━
------------------------------------------------------------
error: reference failed, rustc_parse passed
test: rust/tests/ui/parser/forced-keywords/pre-2021-edition-pass.rs
reference error: invalid token RESERVED_TOKEN at `k#fn }
//~^ WARNING parsed as a forced keyword in Rust 2021 and onward
//~| WARNING this changes mea…`
10: ensure! { k#fn }
              ━
------------------------------------------------------------
passed: 78079
failed: 7
finished in 3 minutes 25 seconds

Fixing the actual check will require more extensive changes to the actual reference grammar that should be done later

[fmease]

Fixing the actual check will require more extensive changes to the actual reference grammar that should be done later [emphasis: mine]

Note that this feature is perma-unstable and should not be observable on stable. The Reference should not mention this token kind.

From the perspective of the Reference k#ident continues to be treated as a reserved prefixed ident in Rust >=2021 just like e.g., prefix#ident.

It just "happens to be case" that rustc emits a different diagnostic for k#ident compared to other reserved prefixes, namely something like "forced keywords are unstable".

What happens if the source code also contains #![feature(forced_keywords)] is out of scope for the Reference / not under its purview.

[traviscross]

Thanks @DanielEScherzer.