(⚠️ Note: This PR requires #1522 to be merged before it can be tested)

These three cases were not recognized by the format specifier parser:

• The nvo, na, nr and nd specifiers may appear together with another specifier, e.g. nvoXb is a valid format specifier.
• Arrays with static size specifiers may contain additional format specifiers, e.g. ptr,[10]s8 is a valid expression.
• Arrays with dynamic size specifiers may contain full C++ expressions in them, e.g. ptr,[mLength-mFree] is valid.

Without this change, the full expression e.g. (foo).bar,nvoXb would be passed to LLDB, which obviously yields a compilation error.

None of these three cases can be implemented easily with the tools we currently have available (as far as I'm aware), but by ensuring we at least recognize and strip such format specifiers, Natvis does not break completely.

Min repro

Say we want to implement a custom UUID type, where we want to use Natvis to render a 16-byte array as a nice string:

struct UUID
{
	union
	{
		unsigned char mData[16];
		unsigned long long mDataWide[2];
	};
};

int main(int argc, char* argv[])
{
	UUID uuid;
	uuid.mDataWide[0] = 0x1234567898765432;
	uuid.mDataWide[1] = 0xABCDEF9FEDCBAF00;

	__builtin_debugtrap();
}

Natvis:

<Type Name="UUID">
	<DisplayString>UUID {mData[0],nvoXb}{mData[1],nvoXb}{mData[2],nvoXb}{mData[3],nvoXb}-{mData[4],nvoXb}{mData[5],nvoXb}-{mData[6],nvoXb}{mData[7],nvoXb}-{mData[8],nvoXb}{mData[9],nvoXb}-{mData[10],nvoXb}{mData[11],nvoXb}{mData[12],nvoXb}{mData[13],nvoXb}{mData[14],nvoXb}{mData[15],nvoXb}</DisplayString>
</Type>

Before

After