An all-caps word containing ß is split into bogus words, so ToMacroCase isn't idempotent: "straße" → "STRAßE" → "STR_Aß_E" #87

What's wrong

ToUpperInvariant('ß') returns ß, and so do other lowercase letters with no single-character uppercase form, such as fi and ʼn. As a result ToMacroCase can produce words like STRAßE that still contain a lowercase (Ll) letter. When text like that is converted again:

• IsWordBoundary (CaseConverter/CaseConverter.cs:118-145) treats ß as a lowercase letter. The acronym-tail rule (upper, upper, then lower) inserts a break before A, and the lower→upper rule inserts a break before the E that follows ß.
• IsAllCaps (:281) returns false for STRAßE, so ToTitleCase doesn't normalize it as an all-caps word.

Repro (net10.0, at 0c1a1d0)

"straße".ToMacroCase();                        // "STRAßE"
"straße".ToMacroCase().ToMacroCase();          // "STR_Aß_E"        expected "STRAßE"
"STRAßE".ToSnakeCase();                        // "str_aß_e"        expected "straße"
"STRAßE".ToPascalCase();                       // "StrAßE"          expected "Straße"
"STRAßE".ToTitleCase();                        // "Str Aß E"        expected "Straße"
"MAßNAHME".ToSnakeCase();                      // "m_aß_nahme"      expected "maßnahme"
"MAX_GRÖßE".ToCamelCase();                     // "maxGrÖßE"        expected "maxGröße"
"maßnahmeLimit".ToMacroCase().ToSnakeCase();   // "m_aß_nahme_limit" expected "maßnahme_limit"
"file".ToMacroCase().ToMacroCase();             // "fi_LE"            expected "fiLE"

Why it matters

German identifiers and config keys (MAX_GRÖßE, STRAßE) are ordinary input. Any pipeline that converts twice, or reads back a macro-case constant it generated itself, gets names split in the wrong places. The library's own output isn't a fixed point of its own conversion.

Suggested fix / acceptance criteria

• Treat a letter that uppercasing leaves unchanged as caseless for word-boundary purposes, for example IsCasedLower(c) => char.IsLower(c) && char.ToUpperInvariant(c) != c. Use it in place of char.IsLower in the acronym-tail rule (:125) and in the lower→upper rule (:132).
• IsAllCaps skips letters that have no uppercase mapping.
• Add regression tests for the examples above, including x.ToMacroCase().ToMacroCase() == x.ToMacroCase().

This is distinct from #84, which is about the upper→lower round trip changing lowercase letters (final sigma, micro sign).