fix(regex): a repeat count over RE_DUP_MAX is "Invalid content of \{\}"

A `\{n\}` / `\{n,m\}` count exceeding RE_DUP_MAX (0xffff) reached the interval
compiler and failed the i16 counter conversion with "Regular expression too
big". GNU rejects it earlier, while reading the number (GET_UNSIGNED_NUMBER in
regex-emacs.c), with `(invalid-regexp "Invalid content of \{\}")`. Add the
RE_DUP_MAX bound check right after parsing the bounds so an over-large count
signals the same error as GNU.

Fixes div_r5_regex_repeat_interval_edge_validation. (Counts in 32768..=65535,
which GNU compiles but neomacs's i16 counters still reject, remain a separate
counter-width follow-up: div_r6.)

Author: eval-exec

neovm-core/src/emacs_core/regex_emacs.rs

@@ -2295,6 +2295,16 @@ fn parse_interval(
         Some(min) // \{n\} — exact count
     };
 
+    // GNU's GET_UNSIGNED_NUMBER (regex-emacs.c) rejects a repeat count that
+    // exceeds RE_DUP_MAX (0xffff) with `(invalid-regexp "Invalid content of
+    // \\{\\}")` while reading the number -- not the generic "too big".
+    const RE_DUP_MAX: usize = 0xffff;
+    if min > RE_DUP_MAX || max.is_some_and(|m| m > RE_DUP_MAX) {
+        return Err(RegexCompileError {
+            message: "Invalid content of \\{\\}".to_string(),
+        });
+    }
+
     // GNU regex-emacs.c:2390 rejects a descending interval where a finite
     // upper bound is smaller than the lower bound (e.g. `a\{2,1\}`), signaling
     // `(invalid-regexp "Invalid content of \\{\\}")`.  An unbounded `\{n,\}`