[libTooling] Fix use of `char` in comparison.

[ymandel]

Summary

Fixes Transformer's Range parser to handle char in a platform-independent way.

[lbenes]

I can confirm that this commit fixes the Wtype-limits warning when built on Raspbian armhf.

clang/lib/Tooling/Transformer/Parsing.cpp

 
 // Returns a new ParseState that advances \c S by \c N characters.
 static ParseState advance(ParseState S, size_t N) {
   S.Input = S.Input.drop_front(N);
   return S;
 }
 
 static StringRef consumeWhitespace(StringRef S) {
-  return S.drop_while([](char c) { return c >= 0 && isWhitespace(c); });
+  return S.drop_while([](char c) { return isASCII(c) && isWhitespace(c); });
 }
 
 // Parses a single expected character \c c from \c State, skipping preceding
 // whitespace.  Error if the expected character isn't found.
 static ExpectedProgress<llvm::NoneType> parseChar(char c, ParseState State) {
   State.Input = consumeWhitespace(State.Input);
   if (State.Input.empty() || State.Input.front() != c)
     return makeParseError(State,
                           ("expected char not found: " + llvm::Twine(c)).str());
   return makeParseProgress(advance(State, 1), llvm::None);
 }
 
 // Parses an identitifer "token" -- handles preceding whitespace.
 static ExpectedProgress<std::string> parseId(ParseState State) {
   State.Input = consumeWhitespace(State.Input);
   auto Id = State.Input.take_while(
-      [](char c) { return c >= 0 && isIdentifierBody(c); });
+      [](char c) { return isASCII(c) && isIdentifierBody(c); });
   if (Id.empty())
     return makeParseError(State, "failed to parse name");
   return makeParseProgress(advance(State, Id.size()), Id.str());
 }
 
 // For consistency with the AST matcher parser and C++ code, node ids are
 // written as strings. However, we do not support escaping in the string.
 static ExpectedProgress<std::string> parseStringId(ParseState State) {