fix(codegen): sanitise XSD enum values for valid Go identifiers

[tnymlr]

Problem

toGoName() only splits on _, -, and ., so XSD enumeration values containing spaces, slashes, colons, plus signs, or asterisks produce invalid Go identifiers in generated const names.

Examples of values that break code generation:

• "EIR Sync Error" → ExceptionTypeEIR Sync Error (space)
• "application/xml" → MediaTypeCodeApplication/xml (slash)
• "application/atom+xml" → MediaTypeCodeApplication/atom+xml (slash, plus)
• "*/*" → MediaTypeCode*/* (asterisk, slash)
• "1:M" → ExceptionType1:M (colon, digit-start)

This causes go/format to reject the generated source with errors like:

unparsable Go source: 367:30: expected ';', found ':'

Fix

Change toGoName() to split on any character that is not a letter or digit (!unicode.IsLetter(r) && !unicode.IsDigit(r)). This handles all non-identifier characters uniformly.

Additional changes:

• Values that produce digit-leading identifiers are prefixed with V (e.g. "1.0" → V10)
• Values that are entirely non-alphanumeric produce "Value" as a fallback
• Original casing within each word part is preserved to maintain existing camelCase names like GetWeather and DownloadRequest

Test plan

• Added TestToGoName with 30 test cases covering spaces, slashes, colons, plus signs, asterisks, MIME types, charset names, and digit-leading values
• Updated golden files for the two cases affected by the change (digit-prefixed enum consts, preserved camelCase in inline type names)
• Full test suite passes
• Tested against real-world Optus wholesale WSDL files containing these enum patterns