Fix panic in unit parsing with non-UTF8 code (#16355)

# Description
Trying to parse non-UTF8 data as a value with unit (part of every
literal parse) introduced a replacement character which shifted the span
information so the indices where incorrect and triggered
a panic.

This has been resolved by removing a bad `String::from_utf8_lossy`

# User-Facing Changes
One less possible panic

# Tests + Formatting
Added a test with the original reproducer from fuzzing:

File with `0\xffB` where the `\xff` represents the non utf-8 char `FF`
run as a script to trigger

crates/nu-parser/tests/test_parser.rs

@@ -292,6 +292,21 @@ pub fn parse_filesize() {
     assert_eq!(value.unit.item, Unit::Filesize(FilesizeUnit::B));
 }
 
+#[test]
+pub fn parse_non_utf8_fails() {
+    let engine_state = EngineState::new();
+    let mut working_set = StateWorkingSet::new(&engine_state);
+
+    // Panic when parsing units was triggered by non-UTF8 characters
+    // due to bad handling via `String::from_utf8_lossy`
+    //
+    // See https://github.com/nushell/nushell/pull/16355
+    let _block = parse(&mut working_set, None, b"0\xffB", true);
+
+    // Asserting on the exact error doesn't make as much sense as
+    assert!(!working_set.parse_errors.is_empty());
+}
+
 #[test]
 pub fn parse_cell_path() {
     let engine_state = EngineState::new();